Corrections Log

Michael Whitman · Trading since 2017, full-time since 2020 · YouTube creator, 137K subscribers · Published 2026-09-23 · Updated 2026-10-28

This corrections log lists the factual corrections made to this site since the log began on 22 September 2026: the date it was corrected, the page, what the page said, and what it says now. Most were found by the site's own recompute checks rather than by readers.

How this log works

When a figure, a worked example or a claim on this site turns out to be wrong, the page is fixed and the fix is listed here. Each row gives the date the correction went live, the page, what it said before, and what it says now. Spelling, wording and layout changes are not listed; anything that changed a number, a calculation or a factual claim is.

Most of the errors below were caught by the site’s own checks: every worked example is recomputed by a script before a chart of it is drawn, and a chart cannot be drawn from a number that does not recompute. That process is described in the editorial policy.

Found an error? Leave a comment on any of my YouTube videos naming the page. I read them, and a confirmed error is fixed and added here.

Corrections

28 October 2026

Page	What was wrong	What it says now
How to Read Time and Sales	The page called time and sales “the only display” that shows what participants actually did.	It is the one display that lists each trade participants actually made; price charts and volume also show what traded, in summary.
Reversals	The worked example’s prices (higher lows of 99.60, 100.03 and 100.06, a lower high of 100.71) came from an invented chart, not a real market.	The example now uses real SPY prices: a lower high of $775.14 on 22 September 2026 and a confirming close of $764.20 on 29 September.
Order Block	The page said a zone is “mostly hollow” by the third time price returns to it.	On QQQ’s daily chart since 1999, later returns held no less often: 40.27% of first returns, 41.44% of second and 46.34% of third.

27 October 2026

Page	What was wrong	What it says now
CCI Indicator	The page said the commodity channel index “has no ceiling” and is unbounded.	Its formula does have a limit, ±666.67 at the usual 20-bar length (the bar count divided by 0.03); on SPY daily bars from 1993 to October 2026 its most extreme reading was -424.10 (24 August 2015).
How to Use the CCI	The page called the CCI unbounded with no upper limit.	Its ceiling is distant rather than absent: ±666.67 at 20 bars, so a reading past +100 or +200 still does not mean the move is done.
CCI vs Williams %R	The page said the CCI has no ceiling.	The CCI’s formula caps a 20-period reading at ±666.67, far beyond its ±100 lines.
RSI vs CCI	The page said the CCI “is not capped at all” and “can print any value”.	The CCI can run to ±666.67 at 20 bars, far past its usual lines, against 0 to 100 for RSI.
CCI vs Awesome Oscillator	The page said neither indicator has a ceiling.	Only the awesome oscillator is open-ended; the CCI’s limit is ±666.67 on its 20-bar setting.
CCI vs MACD	The page said neither has a fixed ceiling.	MACD has none; the CCI tops out at ±666.67 with 20 bars.
CCI vs Money Flow Index	The page said the CCI has no ceiling and can print any value.	The CCI’s formula allows readings out to ±666.67 at 20 periods; on SPY it reached 310.2.
Stochastic RSI vs CCI	The page said the CCI has no upper limit.	On the usual 20-bar setting its math limit is ±666.67.
Stochastic vs CCI	The page said the CCI reading has no upper or lower limit.	Its limits sit far out, at ±666.67 when it looks back 20 bars.
Bid-Ask Spread	The page said an odd lot is fewer than 100 shares.	Since November 2025 a round lot is 100 shares only for stocks priced at $250 or less; from $250.01 to $1,000 it is 40 shares (SEC Release No. 34-104209).
Pattern day trader rule (8 pages: meme stock, settlement, trader tax status, high-frequency trading, full-time trading, trading capital, pattern day trader, margin call calculator)	The pages said FINRA’s pattern day trader rule ended on 4 June 2026 without saying a broker can keep applying it during the phase-in.	FINRA Regulatory Notice 26-10 replaced the rule with intraday margin standards effective 4 June 2026, and firms may phase the change in until 20 October 2027, so some brokers still apply the old count. The pages now say so and cite the notice and the SEC’s approval order (Release No. 34-105226).
Margin Account	The page said firms had until 20 October 2027 to phase in the new intraday standards, implying all of them would take that long.	Firms that need more time may phase them in until 20 October 2027, so a broker may still be running the old rules; the notice also removed the $25,000 minimum equity.
Fibonacci Retracement	The page said no part of a pullback is more than about 12% of the swing away from one of the five Fibonacci lines.	That is false near the ends of the swing: a point near 0% is 23.6 points of the swing from the nearest line, and one near 100% is 21.4 points away. Between the 23.6% and 78.6% lines no point is more than 8.4% of the swing from a line, and the page now says that.
Settlement	The free-riding paragraph described buying with unsettled funds and selling before the original sale settles, which is a good-faith violation, and said only a pattern of it brings the 90-day restriction.	Free-riding is paying for a purchase only with the proceeds of selling those same shares. Under Regulation T 220.8(c), selling a security before paying for it in full withdraws the privilege of delaying payment for 90 calendar days unless full payment arrives within the payment period.
Shareholders’ Equity	The buyback example said return on capital employed would have fallen.	It would not have moved: $200 million more debt and $200 million less equity leave debt plus equity unchanged, and operating income is unchanged too.

5 October 2026

Page	What was wrong	What it says now
Fair Value Gap vs Order Block	The page said “a gap inside a block is one zone” and that the two “frequently overlap by construction”.	By construction they cannot overlap: a bullish gap’s band starts at candle one’s high, so when candle one is the order block the gap sits directly above it. On IWM since 2000, candle one closed against the move in 351 of 1,194 bullish gaps, and only 67 also came from a move that broke structure.
Fair Value Gap vs Change of Character	The verdict said the change-of-character warning “is right far less often than it fires”.	On the site’s own rule the next structure break confirms or reverses it about evenly: on SPY 110 of 220 were reversed (terms/change-of-character), and on bitcoin since 2014 the next break went the same way 49 times and reversed 38.
FVG vs Imbalance	The page called imbalance a broader idea “with no fixed definition” and said most things called imbalances are not fair value gaps.	This site defines both terms as the same three-candle wick gap (terms/imbalance), and the page now says so. The only difference is a looser two-candle body-gap usage, which on Apple’s daily chart from 2010 found 1,382 bands, 554 of them outside every fair value gap.

4 October 2026

Page	What was wrong	What it says now
CCI vs Money Flow Index	The page said that in a strong move the money flow index “reaches its cap and stays there”.	On SPY daily bars from 1994 to 2026 the money flow index never reached 100; its highest reading was 95.33 (12 November 2004), and its 98 runs above 80 lasted a median of 2 sessions.
Change of Character vs Supply and Demand	The page said “most changes of character fail”.	The site’s own change of character page found the next break reversed exactly half of SPY’s 220 changes of character (110 of 220: 68 of the 110 downward ones and 42 of the 110 upward ones). The page no longer says most fail; it reports that 107 of the 220 break days moved less than one average true range.
Change of Character vs Support and Resistance	The page said “most changes of character fail”.	Same correction: half were reversed by the next break, not most. The page now reports that price came back to the broken level after 196 of 220 (89.09%) and held on 99 of those returns (50.51%).
EMA vs Hull Moving Average	The page said the hull’s reduced lag “is paid for in false turns” and that it “turns early at things that are not turns”.	On SPY daily closes from 1994 to 2026 the 20-day hull turned 915 times against 1,057 for the 20-day exponential average, and only 20.13% of its turns came within 3 sessions of the one before (51.33% for the EMA). The cost of its lower lag is overshoot instead: it sat above every one of its last 20 closes on 720 sessions, which the EMA never did.
EMA vs MACD	The page said MACD “lags slightly more, because it smooths a quantity that was already smoothed”.	The MACD line is a plain difference of two exponential averages with no extra smoothing; on the S&P 500 since 1958 all 549 of its zero-line crosses fell on the same sessions as the 12/26 average crosses. Only the signal line adds a further average, and its same-direction cross came a median 6 sessions before each zero-line cross.
Breakout vs Reversal	The page said reversals “win often and small, because most tests do hold”.	On SPY daily bars from 1994 to 2026, price reached a 60-session high 1,268 times and closed above it on 697 (54.97%), so most tests did not hold. After either close, a 2-ATR drop came first about as often as on any session (35.01% after a close above, 35.03% after a rejection, 36.67% overall).
Breaker Block vs Supply and Demand	The page said the two areas have “opposite histories” and “are traded in opposite directions”.	Both are traded the same way on the return. On SPY from 1996, 45 of 365 structure breaks were also sharp departures, and in 18 of those the supply or demand zone and the breaker were the same candle.
Breaker Block vs Mitigation Block	The page said the mitigation framing “usually expects the return to be sold into rather than bought”, contradicting the site’s own mitigation block definition (a return and then a continuation in the original direction).	Both are traded in the direction of the move before the return; the difference is timing. On SPY 81 breakers later fit the mitigation shape, a median 7 sessions after the break.
Ichimoku vs Parabolic SAR	The page said that when the cloud goes flat “the dots flip constantly” and “every few bars”, so the cloud tells you when the dots’ flips mean nothing.	On SPY daily bars from 1994 to 2026 the parabolic SAR reversed at about the same rate wherever price sat: 10.01 reversals per 100 sessions above the cloud, 8.66 inside it and 9.35 below it.
Distribution	The page called a failed push above the high “the classic mark” of distribution.	On SPY from 1994 to 2026, tight ranges near a 52-week high that had a failed push above their top broke downward 10 times in 73, against 54 in 163 without one; among ranges whose close sat in the top third, 9 of 63 against 18 of 115.
Weighted Moving Average vs Ichimoku	The page’s data section cited the site’s own illustrative measurements as its source.	The page now carries no statistical study and says so; it works one real SPY example (14 and 17 August 2026) instead.
Breaker Block vs Change of Character	The page said requiring a change of character before a breaker “is a real filter”.	On SPY daily bars from 1996 to 2026, breakers left by a change of character held on 41.41% of returns against 39.67% for breakers left by a continuation break (p = 0.80); the page now says the filter gives fewer setups, not better ones.
Fundamental Analysis vs Elliott Wave	The page called January 1973 to July 1980 (7.5 years) the S&P 500’s “longest wait for a new high”. Counting the record back to the 16 September 1929 close, the longest wait ran from September 1929 to September 1954; 1973 to 1980 is the longest since 1950.	The sentence is removed (it was not specific to this comparison).
Candlesticks vs Heikin-Ashi	The page said heikin-ashi bars do not show real prices, “so every level you read off one is wrong”, and that the heikin-ashi open “never happened”.	On SPY daily bars from 1996 to 2 October 2026 the heikin-ashi high and low were the real traded high and low on 3,410 of 7,738 sessions (44.07%), and the heikin-ashi close always lies inside the day’s range; the heikin-ashi open fell outside the day’s range on the other 55.93%.
What Is the Bond Market?	The page said the bond market is “larger than the stock market”.	SIFMA puts US bonds outstanding at $61.2 trillion and the market value of US-listed companies at $68.9 trillion for 2025, so the US bond market was about 11% smaller; worldwide, bonds ($160.7 trillion) were slightly larger than stocks ($157.8 trillion).
Bollinger Bands vs Donchian Channels	The page said “a fifty-five-bar breakout is much rarer than a twenty-bar one”.	On SPY daily bars from 1994 to 2 October 2026 the high topped the prior 55-day high on 1,318 sessions (15.99%) and the prior 20-day high on 1,720 (20.87%).
Bollinger Bands vs Donchian Channels	The page said “longer lookbacks produced fewer and better breaks”.	On SPY from 1994 to 2026, the close 10 sessions later was higher after 62.37% of new 55-day highs and 63.02% of new 20-day highs, so the longer lookback was not better.
Bollinger Bands vs Donchian Channels	The page said a Donchian edge “sits still until a new extreme is made, then steps”.	The upper 20-day line also falls when an old high leaves the window; on SPY that happened on 698 of 8,243 sessions (8.47%).
Accumulation Distribution vs OBV	The page said the two lines “will point the same way almost always”.	On SPY from 1994 to 2026 they moved the same way on 80.04% of days and in 290 of 412 separate 20-session blocks (70.39%).
EMA vs WMA	The page said the two lines track each other “closely enough that no method would produce different trades”, and that running both is “the same line drawn twice”.	Measured on S&P 500 closes from 1958 to 2 October 2026, price closed between the two 20-day lines on 7.87% of sessions (1,362 of 17,304), where a rule that reads “above the line” gives two answers. At 502 turning points price crossed the weighted line first 186 times, both on the same session 296 times, and the exponential line first 20 times.
Stochastic RSI vs Williams %R	The page said not to run both because they are “near enough the same measurement”, “one reading drawn twice”. That describes Williams %R and the plain fast stochastic, not the stochastic RSI.	On SPY daily bars from 1996 to 2 October 2026, the daily changes in Williams %R and stochastic RSI %K had a correlation of 0.50: related, not one reading.
Stochastic RSI vs Williams %R	The page said both indicators pin: “in a sustained move each reaches its extreme and stays”.	On the same SPY data Williams %R sat at exactly 0 or -100 on 1.11% of sessions; the raw stochastic RSI sat at exactly 0 or 100 on 29.53%.
Margin account	The page said “exactly one” video in the 24,971-video study covers margin accounts, with the highest median views of any market-mechanics topic.	3 titles name margin accounts, margin trading or buying on margin, at a median of 125,237 views. The “highest” comparison was never measured and is removed.
Margin account	The page said a Regulation T call “cannot be waived by anyone at the firm”.	Under 12 CFR 220.4 the call must be met within one payment period, which the firm’s examining authority can extend on the firm’s application, and a deficiency of $1,000 or less needs no action.
BOS vs CHoCH	The page said continuation is “the ordinary case”, that the first break against a run “is right occasionally”, and that it “is the market pausing far more often than it is the market turning”.	On SPY daily bars from 1993 to 2026 with five-bar swings, about half of the 219 reads that a change of character opened were followed by a break of structure in the new direction (109 of 219); 42 of 110 upward reads and 68 of 109 downward reads got none.
BOS vs CHoCH	The page said a marginal break should not be treated as decisive: “Displacement is the filter.”	Changes of character that closed further past their level than the median (0.465%) were followed by a break of structure 48 times in 109, against 61 in 110 for the smaller ones (Fisher exact p = 0.11): a wider close was not a better guide on this data.
Bollinger Bands vs Keltner Channels	The page said the two “agree almost always by construction and the second one adds no information”.	On SPY daily bars from 1994 to 2 October 2026, the close was outside one envelope and inside the other on 1,217 of 8,243 sessions (14.76%).
Retained earnings	The worked example said paid-in capital of 1,200 and retained earnings of 2,286 made total equity funding of 3,230, 63% of it from operations. 1,200 + 2,286 is 3,486, and 2,286 is 65.58% of that.	The example now uses $1,200 + $2,030 = $3,230, of which 62.85% came from the company’s own operations.
Retained earnings	The page said “a company cannot pay a dividend larger than its retained earnings in most jurisdictions”. Retained earnings is not the legal limit: Apple paid dividends while carrying an accumulated deficit.	The page says the legal test is set by the law of the state of incorporation, and that Apple paid $15,421 million of dividends in its 2025 fiscal year with an accumulated deficit at both ends of it (-$19,154 million at 28 Sep 2024, -$14,264 million at 27 Sep 2025, per its 10-K filings).
Paid-in capital	The page said the balance “moves only on a corporate action”, that “for most established companies the balance sits unchanged for years”, and that buybacks appear as treasury stock. Paid-in capital rose in each of the last ten fiscal years at all five large companies measured, and Apple, Microsoft and Tesla report no treasury stock at all.	The page measures it in 10-K filings: Amazon’s additional paid-in capital rose $126,630 million over 2016-2025 against $126,559 million of stock-based pay, and only Amazon and Coca-Cola of the five report treasury stock.
Intraday trading	The page said that dividing your round-trip cost by the average bar range and multiplying by trades a year gives “the percentage of range you must capture annually just to break even”. That calculation gives a number of bars’ worth of movement, not a percentage.	The page works the example through on SPY: a hypothetical $0.02 round trip was 3.92% of the median five-minute bar from 10 July to 2 October 2026, so 1,008 round trips a year hand over about 39.5 median bars of movement, while the same $0.02 was 0.38% of a median day.
Zero-sum game	The page said half of all participants must do worse than the average result, that in a zero-sum market “somebody must” win, and that every stop filled is somebody’s entry. Half do worse than the median, not the mean; after costs every participant in a period can finish behind; and a filled stop can be matched against another trader’s exit.	The page says the group’s average result is a loss equal to the costs and that this says nothing about how many individuals lose, that after costs everyone can finish behind, and that a filled stop meets an order on the other side. It shows from the CFTC’s reports that long and short open interest were equal in 13,995 of 14,055 market-weeks of 2026.
Momentum trading vs scalping	The page called the gap between how popular scalping is and how workable it is “one of the widest gaps between popularity and workability measured anywhere on this site”. Workability was never measured, on this page or any other.	The page measures both styles on SPY: a round trip of about $0.026 a share (an assumed one-cent spread plus the SEC and FINRA sale fees) was 11.83% of a median five-minute move and 0.22% of a median one-month move, and consecutive five-minute moves went the same way 47.41% of the time against 50.00% expected by chance.
Stop loss	The page said 91 of the 147 stop-loss video descriptions in our study (63%) mention invalidation, failure or going wrong, “the highest rate measured anywhere in this glossary”. The count matched the word “loss”, which is in the topic’s own name: 83 of the 91 matched on that word alone, and only 2 descriptions use words such as invalidation, fail, wrong or mistake.	The claim is removed. The page now measures stops on real prices: on SPY since 1996 a 5% stop was hit within 20 sessions after 22.86% of closes, and 14.74% of those hits opened beyond the stop.

3 October 2026

Page	What was wrong	What it says now
Accumulation	The page said volume was “the only genuinely measurable part” of the accumulation framework, that there was “no way to tell in advance” which way a sideways stretch would break, that ranges “are entered slowly and left quickly”, that every chart of a major low has an accumulation range on it, and that in the site’s video study accumulation appears “almost entirely” inside Wyckoff and smart-money material. The fall, width, length and position in the range are measurable too; the speed and major-low claims were never measured; and in the study 51 of the 92 titles that name accumulation also name Wyckoff, smart money, SMC or ICT.	The page fixes a rule and measures 1,515 ranges after a 10% fall in 56 US stocks and ETFs from 2000 to 2026. Where day 30 closed in the range went with which edge broke first: 29.19% of bottom-third ranges left upward first, against 80.90% of top-third ones. Sixty sessions later the ranges were higher 58.49% of the time, against 59.48% for any day. The unmeasured claims and the video-study line are removed.
Stock exchange	The page said an exchange “guarantees a match”, that official volume comes from the exchange, “which is why off-exchange prints arrive late”, and that the volume and prices on a chart all originate on the exchange. A limit order away from the market does not fill. In September 2026, by Cboe’s figures, 50.63% of US shares traded off exchange, and FINRA Rule 6380A requires those trades to be reported no later than 10 seconds after execution.	The page says an exchange matches what can be matched, gives Cboe’s venue shares for September 2026 (exchanges 49.37%, off exchange 50.63%) and for each September since 2021, and quotes FINRA Rule 6380A on the 10-second reporting limit.
Change of character	The page said most breaks against a trend resume it rather than reverse it, that acting on one “costs nothing and is frequently right”, that a “higher-probability version” waits for a retest, that it is “the oldest observation in technical analysis”, that volume is “the only independent check”, that higher timeframes “remove most of them” and the survivors are “the ones worth watching”, and that the site’s video study finds the term “almost entirely inside smart-money-concepts content”. None of these was measured. In the study, 36 of the 100 titles that name change of character or CHoCH also name smart money concepts, smart money or ICT.	The page states its swing setting (five bars each side) and measures all 220 changes of character on SPY’s daily chart from 1993 to 2026: the next break went the other way 110 times out of 220. After a downward one, SPY fell another 10% within 60 sessions 15.60% of the time, against 13.58% after any day, a gap within chance (p = 0.49). The probability, superlative and video-study claims are removed.
False breakout	The page said breaks of minor highs in a strong trend “succeed far more often than they fail”, called thin volume “the most common tell”, said waiting for a retest “removes most of the ambiguity” and a higher timeframe “removes the majority” of false breaks, said stops above the high “were the point” and are “why breaks so often reverse”, and said the liquidity above a high “has been consumed” by the time the break is visible. None of these was measured or sourced.	The page measures 623 closes above a 20-day high on SPY’s daily chart from 1993 to 2026: 343 (55.06%) closed back below the level within five sessions, against 56.93% expected from ordinary days of the same size, a gap within chance (p = 0.33). Volume is described as a filter the page does not test, the stop explanation as “one common explanation”, and the retest, timeframe and liquidity claims are removed or reworded as untested.
Backtesting, debt ratio, forward testing, stocks vs penny stocks	Each called January 1973 to July 1980 the S&P 500’s longest spell below a record. Counting the series back to 1929, the longest ran from September 1929 to September 1954.	The sentence now says the 1973 to 1980 spell is the longest since 1950.
S&P 500	The page said falls of 5% or more happened 68 times since 1950 (0.89 a year), but one of the 68 was already under way in January 1950.	67 falls began after 1950 (0.87 a year), and one more was already under way in January 1950.
S&P 500, buy and hold, methodology and 239 other pages	The pages said the S&P 500 closed below its previous record on 92% of trading days since 1950, and set a record on about 8% of days (about one in twelve). Some also said it had 74 separate falls of 5% or more below a record, about 0.96 a year. The count started at the first close of 1950 and missed the September 1929 high, which the index did not pass again until September 1954, so it counted closes from 1950 to 1954 as records when they were not.	93% of trading days closed below the previous record (92.95%, Yahoo Finance daily closes, 3 Jan 1950 to 18 Sep 2026), about 7% set one (about one in fourteen), and there were 68 separate falls of 5% or more, about 0.89 a year. The longest wait (7.5 years, 1973 to 1980) and the worst fall (56.8%, 2007 to 2009) did not change. The data file and the charts that drew the old figure were redone.
Implied volatility	The page said implied volatility “collapses” the moment a result is known and called that drop “the most reliable event in the options calendar”. Nothing on the page measured it, and no data on this site supports the superlative.	Once a result is known, implied volatility usually drops, and the page links to the IV crush measurement: Cboe’s 30-day volatility index for the stock fell after 246 of 251 earnings releases from Apple, Amazon, Alphabet and IBM. The page now also carries real Cboe option quotes from 25 September 2026 in place of an illustrative diagram.
Break of structure	The page said a “higher-probability version” waits for price to return to the broken level and hold, said the definition needs “no settings”, called it “the oldest definition in technical analysis”, said traders watching the same price is “a real effect”, and said the site’s video study covers the term “almost entirely inside smart-money-concepts material”. None of these was measured. In the study, 31 of the 99 titles that name break of structure also name smart money concepts or ICT. A swing point needs a setting: how many bars on each side it must beat.	The page states its swing setting (five bars each side) and measures every break on SPY’s daily chart from 1993 to 2026: 66.90% of 142 upward breaks were higher 20 sessions later, against 63.91% of all days, a gap within chance (p = 0.49). Dow theory, from Charles Dow’s editorials before 1902, is given as the older idea; the “real effect” line now reads as a condition (“if the level matters at all”); the probability and video-study claims are removed.

26 September 2026

Page	What was wrong	What it says now
Volume flow indicator	The page said that “in a measured sample the bar at the ninetieth percentile is 6 times the one at the tenth”. No data behind the page held a bar-size distribution; 6 was the ratio of the page’s own made-up example (1.20 against 0.20).	Measured on SPY daily bars since 1993, the ninetieth-percentile move (1.43%) is about 19 times the tenth (0.07%), and about 12 times over the last 252 sessions.
Candlesticks	The page referred to “those 253 descriptions” that never mention a failed read. 253 does not match any count in the video study.	308: the study has 312 descriptions for its 645 candlestick titles, and 4 of them mention a failed read.
Chart patterns, market trend	Chart patterns called continuation shapes “the more reliable group” and said reversal shapes fail more often. Market trend said trend-following systems “lose most of their trades”. No data on this site supports either claim.	Chart patterns says the case usually made for continuation shapes is that they agree with what is already happening, and that a reversal needs more to go right. Market trend says trend-following systems are built to take many small losses.
WaveTrend	The page called the volume flow indicator “another LazyBear tool”.	The volume flow indicator is Markos Katsanos’s, first published in June 2004; LazyBear published the common TradingView version.
Breaker block vs mitigation block, hull moving average comparisons, order-type comparisons and 18 other comparison pages (21 in all)	A recount of the 24,971-video YouTube corpus found 34 stale or mis-added figures and phrases. Where two subjects were counted together, the two counts had been added, so titles that name both were counted twice (for example 64 instead of 55). Some counts predated alias matching (break of structure 51, now 110; hull moving average 83, now 101), and several pages called a pair “the least-covered” or “the thinnest” on the site when other pairs had fewer videos.	Each total now counts every video once and says how many titles name both subjects. The updated counts come from one script run over the deduplicated corpus, and each was checked a second time with a separate script. Overstated superlatives now read “one of the” or describe the pair as thinly covered.
Short selling	The short-sale price test was said to switch on when a stock “falls more than 10% in a day”. SEC Rule 201 triggers on a fall of 10% or more from the prior day’s closing price, so a fall of exactly 10% counts (AMD on 7 December 2018: prior close $21.30, low $19.17) and the fall is measured from the previous close, not across the day.	A stock that falls 10% or more from the previous day’s close triggers the alternative uptick restriction for the rest of that session and the next. The rule is explained in full on the new short-sale restriction page.
EMA vs weighted moving average, SMA vs weighted moving average, hull vs weighted moving average and the three pages comparing the weighted moving average with ichimoku, parabolic SAR and supertrend	The weighted moving average was counted in 11 titles at a median of 201 views across 11 channels. The search for “weighted moving average” also matched 3 titles about the volume-weighted moving average, which is a different indicator. That changed the ratios drawn from it (for example “twelve times the videos” beside supertrend), the hull page’s count of titles naming both (2, now 1), and the ichimoku page’s lesson share (45%), under-50,000 share (82%) and median length (6.2 minutes). The EMA page also said “hundreds” of exponential-average titles, where the count is 198.	8 titles, median 197 views, 8 channels, with volume-weighted titles left out. The ratios are recomputed (supertrend now has nearly seventeen times the videos), one title names both the hull and the weighted average, the ichimoku page gives 25% lessons, 88% under 50,000 and 5.6 minutes, and the EMA page says nearly two hundred.
Pattern day trader	A day-trading buying power call was said to restrict the account to cash-available trading “for ninety days until it was met”. Under FINRA Rule 4210 as it stood before the 4 June 2026 change, the ninety-day restriction applied only if the call was still unmet after five business days.	A call still unmet after five business days restricted the account to cash-available trading for ninety days or until it was met.

25 September 2026

Page	What was wrong	What it says now
RSI(2) strategy, three down days, five-day low, ATR profit target, 200-day moving average, overnight risk	The five retest stocks (INTC, PYPL, DIS, NKE, WBA) were described as stocks that “went sideways or down”. On price, Intel rose 168.2% and PayPal 60.9% from January 2015 to August 2026, so that was wrong for two of the five. Walgreens was also described as tested to August 2026, but it stopped trading on 27 August 2025 when it was taken private.	The five are described as stocks that lagged the S&P 500, which rose 276.7% over the same window; all five did. The pages and their charts now note that Walgreens’ runs end in August 2025. The backtest results themselves did not change.

23 September 2026

Page	What was wrong	What it says now
Original research	The video study was said to cover “10,608 channels”. That count mixed channel IDs (present on only 5,849 of the 24,971 rows) with channel names, so a channel seen both with and without an ID was counted twice.	9,458 channels, counted by channel name on every row, the same way every topic page counts.
Every page (the byline)	The byline printed a raw timestamp, “Reviewed 2026-09-06 00:00:00 +0000 UTC”, on every page, and the date given to search engines as “last modified” was the same stored string, not the date the text last changed.	The byline shows “Published” and “Updated” as plain dates, and “Updated” moves only when the page’s text actually changes (a stored fingerprint of each page’s text is compared at every build).

22 September 2026: worked examples that did not recompute

Found while drawing each page’s worked example as a figure: the drawing is generated from the numbers and checked against the page, and these did not match.

Page	What was wrong	What it says now
Crypto DCA calculator	The average price paid was computed as the plain average of the prices, which is not what dollar-cost averaging pays. The example showed an average of 37,500, an end value of 3,120 (+20%), and for the V-shaped path 21,600, 3,616 and +39.1%.	The average is total dollars divided by total coins bought: 36,975, end value 3,164 (+21.7%); V-shaped path 21,754, 3,586 and +37.9%. The calculator itself uses the corrected formula.
How to handle a losing streak	At a 55% hit rate over 100 trades, a run of 6 losses was given a 54.71% chance (“more likely than not”), a run of 8 14.49% and a run of 10 3.05%.	Counted exactly, the same way as the losing-streak calculator: 36.29% (“about one time in three”), 8.37% and 1.71%.
Fractional shares	Buying whole shares only, about $100 was said to sit uninvested on average.	About $300 sits uninvested on average ($200 a month into a $640 share).
How to calculate portfolio beta	After holding A doubles, the portfolio beta was given as about 1.09.	About 1.06.
Concentration ratio	The Herfindahl example gave Market X a score of around 1,700.	About 1,620 if the twenty small firms are equal: 4 x 20 squared + 20 x 1 squared = 1,620.
Convenience yield	The example implied a convenience yield of “about 10”.	About 7 a barrel, roughly 9% of spot (83 - 76 = 7).
Active risk	The gap between a closet indexer and an index tracker was “roughly 15 percentage points”.	About 14 (20.2 - 5.8 = 14.4).
Gross profit	With revenue up 20% and expenses up 5%, the low-margin company’s operating income was said to rise to 58.	About 46 (1,200 x 24% - 241.5 = 46.5).
Trial balance	Four year-end adjustments were said to cut operating profit to 100, “a third” of it.	56, nearly two thirds of it gone (150 - 45 - 30 - 12 - 7 = 56).
Intermarket sweep order	Venues C and D were each said to show 1,000 shares, which made the displayed total 2,600 against a 2,000-share order.	700 each, so the four venues display exactly the 2,000 the order sweeps.
Market manipulation	Being stopped out and buying back a $10,000 position quoted 0.2% wide was said to cost $40.	About $20, one full spread, as the site counts a round trip everywhere else.
Penny stock	The spread was said to be paid on entry and again on exit.	A round trip gives up one full spread (10% at 0.50 bid, 0.55 ask), which matches the page’s next paragraph.
Trading the point of control	43 against 35 on fifty observations was “about one and a half standard errors apart”.	A little under two standard errors apart (0.86 against 0.70 on 50).
Avalanche vs snowball	The saving was “about 4.9% of the interest”.	About 5% (140 / 2,890 = 4.8% and 140 / 2,750 = 5.1%, so neither figure is 4.9% exactly).

22 September 2026: video counts recounted site-wide

Pages quote how many videos in the 24,971-video study name a subject in their title. Until 22 September the count matched raw text, so “ira” was counted inside “miracle” and “art” inside “chart”. Every count was recomputed with one method (whole words, plurals included), which changed 1,933 figures on 775 pages. Most moved slightly. These are the ones where a comparison the page drew from them was wrong:

Page	What was wrong	What it says now
CCI vs MACD	“473 videos on one and 448 on the other”: almost identical coverage.	471 against 67: seven times the coverage for MACD, and more than three times the audience per video for the commodity channel index.
Breaker block vs support and resistance	22 against 145 videos: “seven times the coverage and nearly three times the audience”.	22 against 275: twelve times the coverage and nearly twice the audience.
CHoCH vs MSS	40 against 10 videos: a quarter of the coverage and “roughly ten times the audience per video”.	100 against 12: an eighth of the coverage and roughly four and a half times the audience per video.
Day trading vs momentum trading	1,021 against 82 videos: “nearly three times the audience”.	960 against 82: two and a half times the audience.

Pages that had said no videos existed on a subject now give a positive, recounted figure instead, since a title search that finds nothing proves little about a subject.

What is not in this log

Changes that do not alter a fact: rewording, spelling (the site moved to US spelling throughout on 22 September), new images, new sections and layout. Each page’s “Updated” date shows when its text last changed, and the methodology page explains how figures are sourced.

How pages are researched, checked, reviewed and corrected is set out in the editorial policy. The datasets that every recount above was run against are on the research page, and who is accountable for all of it is on the about page.

This page is educational, not financial advice. Test every idea on your own charts before risking money.