How to audit a trading track record: 7 checks

STS ResearchPublished July 6, 2026Updated November 10, 2026Backtest data as of November 6, 2026

We ran a strict seven-check audit on our own backtested NQ book, 3,139 trades over 15 years, and it passed all seven. But on the check that matters most, the t-stat, only one of our four strategies clears the bar on its own trades. We are telling you that up front, because the point of an audit is to find the thin spot, not to print a row of green checkmarks.

Here are the seven checks. Demand a number for each: a big enough sample, a t-stat over 3, a profit factor above 1 in every era, costs actually subtracted, a drawdown percentage you can trace to a denominator, a planning drawdown from a reshuffle, and the losing years shown. Below is the full scorecard, the threshold for each, and the one line where a strict auditor would push back on us.

Audit scorecard table with seven rows, each a check to demand from any trading track record next to the STS NQ book's own number and grade. Row 1 sample size, bar 200-plus trades multi-year, ours 3,139 trades 2011 to 2026, PASS. Row 2 t-stat edge versus luck, bar t greater than 3, ours clears it for the book but only 1 of 4 strategies clears it on its own trades, PASS as a book. Row 3 per-era profit factor, bar PF greater than 1 in every era, ours 1.05 rising to 1.21 to 1.37 to 1.73, PASS. Row 4 costs included, bar commissions plus slippage in, ours about 4 dollars 10 cents round-turn on every fill, PASS. Row 5 drawdown honesty, bar the percent must state its denominator, ours 25,342 dollars equal to 3.6 percent of net or 3.1 percent of the peak it fell from or 25.3 percent of a 100k account, PASS. Row 6 planning drawdown, bar quoted from a reshuffle not a single path, ours backtested at the 5th percentile, with a reshuffle median of 34,223 dollars and 1 in 20 reshuffles above 51,684 dollars, PASS. Row 7 losing periods shown, bar negative years and months published, ours 3 down years and 67 down months, PASS. Audit scorecard table with seven rows, each a check to demand from any trading track record next to the STS NQ book's own number and grade. Row 1 sample size, bar 200-plus trades multi-year, ours 3,139 trades 2011 to 2026, PASS. Row 2 t-stat edge versus luck, bar t greater than 3, ours clears it for the book but only 1 of 4 strategies clears it on its own trades, PASS as a book. Row 3 per-era profit factor, bar PF greater than 1 in every era, ours 1.05 rising to 1.21 to 1.37 to 1.73, PASS. Row 4 costs included, bar commissions plus slippage in, ours about 4 dollars 10 cents round-turn on every fill, PASS. Row 5 drawdown honesty, bar the percent must state its denominator, ours 25,342 dollars equal to 3.6 percent of net or 3.1 percent of the peak it fell from or 25.3 percent of a 100k account, PASS. Row 6 planning drawdown, bar quoted from a reshuffle not a single path, ours backtested at the 5th percentile, with a reshuffle median of 34,223 dollars and 1 in 20 reshuffles above 51,684 dollars, PASS. Row 7 losing periods shown, bar negative years and months published, ours 3 down years and 67 down months, PASS.
All seven checks pass on 3,139 trades from 2011 to 2026. Check 2 carries a star: the t-stat clears 3.0 as a book, and only 1 of the 4 strategies clears it on its own trades.

Whose track record is this (read this first)

The numbers below come from the STS NQ book, our own backtested record. Four systematic NQ strategies run as one portfolio that holds a single position at a time. TradingView backtests, 2011 to 2026 (data as of November 6, 2026), 1 or 2 NQ minis a trade (2 when daily ranges are small, 1 when they are large; at today's index level every trade since July 2024 has been 1 contract), commissions and slippage included, $696,892 net over 3,139 trades on the NQ mini. The style is momentum and trend continuation, intraday plus one overnight model. Not mean reversion, not scalping.

That matters for how you read this. The seven checks transfer to any record, yours included. The exact numbers do not. When you run these checks on a signal seller, a prop firm demo, or your own backtest, you will get your own answers. We are using our book as the worked example of what each check looks like when someone actually shows their homework.

How many trades does a track record need to be meaningful?

The rough floor is a few hundred trades across more than one market condition, so a bull run, a chop year and at least one drawdown are all in the sample. Our backtested book is 3,139 trades from 2011 to 2026.

Ask how many trades the record covers, over how many years.

A handful of trades can look brilliant on luck alone. The more trades a record has, the harder it is for luck to carry the whole thing. The rough floor is a few hundred trades across more than one market condition, so a bull run, a chop year, and at least one drawdown are all in the sample.

Our backtested book is 3,139 trades from 2011 to 2026. That count reconciles three independent ways in our export: exit rows, entry rows, and the running cumulative column all land on 3,139. Fifteen years covers two bear markets, a crash, and several flat stretches. This check is the easy one. If a seller can only show you last quarter, stop there.

What t-stat does a trading strategy need to count as real?

Harvey, Liu and Zhu argue for t over 3, not the old bar of 2.0. Our book clears the 3.0 bar, but of our four strategies on their own trades only the Trend sleeve, at 3.15, passes.

Ask for the t-stat. It is a single number that says how far a record sits above pure luck.

The t-stat measures the average trade against how noisy the trades are. A high number means the profit is unlikely to be a fluke. Harvey, Liu and Zhu studied thousands of published strategies and argued the old bar of 2.0 was far too soft once you count how many things people test before they publish a winner. Their tougher line is t over 3.

Our book clears the 3.0 bar comfortably. On a dollar-PnL basis it clears by even more. But here is the honest part almost no vendor shows you: run the same test on each of the four strategies, on its own trades, and only one clears the bar.

The per-strategy t-stats are 3.15, 2.09, 2.56 and 2.93. Only the first, our Trend sleeve at 3.15, passes 3 on its own trades. The other three are soft on their own. We keep them because they run together, one position at a time, and together they lift the combined book well clear of the bar. A strict auditor applying this check strategy by strategy would flag all three, Opening Range Break at 2.09 most of all. We would rather you see that now than find it later.

Bar chart of t-statistic per strategy against a t equals 3 threshold line. Trend sleeve 3.15, a green bar above the line, the only strategy that clears on its own trades. Three red bars below the line: Opening Range Break at 2.09, VWAP Short at 2.56, Overnight at 2.93. A tall blue bar on the right shows the combined book well above the line, labeled all four combined. An amber dashed line marks the t equals 3 Harvey-Liu-Zhu bar. Bar chart of t-statistic per strategy against a t equals 3 threshold line. Trend sleeve 3.15, a green bar above the line, the only strategy that clears on its own trades. Three red bars below the line: Opening Range Break at 2.09, VWAP Short at 2.56, Overnight at 2.93. A tall blue bar on the right shows the combined book well above the line, labeled all four combined. An amber dashed line marks the t equals 3 Harvey-Liu-Zhu bar.
Only the Trend sleeve (3.15) clears t > 3 on its own trades; three strategies sit below the bar. Run one position at a time, the four combine to clear the bar for the book. This is the one place a strict auditor would push back, and we show it rather than hide it.

Here is why the t-stat is the check that matters, not the friendlier ones. Every one of our four sleeves makes money on its own trades, with profit factors from 1.35 to 1.69. Judge them on profit factor and you would call all four winners. Judge them on the t-stat and three of the four fall short. Profit factor tells you a strategy made money in the past. The t-stat tells you whether that money is likely to be edge instead of luck. A record that leads with profit factor while staying quiet about the t-stat is showing you the softer number.

Two side-by-side bar lanes, one row per NQ sub-strategy, read across. Left lane is profit factor, every bar green and past the PF equals 1 line: Trend 1.46, Opening Range Break 1.35, VWAP Short 1.69, Overnight 1.40. Right lane is t-statistic against an amber t equals 3 gate: only Trend clears it at 3.15 as a green bar; Opening Range Break 2.09, VWAP Short 2.56 and Overnight 2.93 fall short as red bars. Same four sleeves, graded winner by profit factor and luck by the t-stat, so profit factor calls all four winners while the t-stat clears only one. Two side-by-side bar lanes, one row per NQ sub-strategy, read across. Left lane is profit factor, every bar green and past the PF equals 1 line: Trend 1.46, Opening Range Break 1.35, VWAP Short 1.69, Overnight 1.40. Right lane is t-statistic against an amber t equals 3 gate: only Trend clears it at 3.15 as a green bar; Opening Range Break 2.09, VWAP Short 2.56 and Overnight 2.93 fall short as red bars. Same four sleeves, graded winner by profit factor and luck by the t-stat, so profit factor calls all four winners while the t-stat clears only one.
Every sleeve clears PF > 1 (green bars, 1.35 to 1.69), so profit factor calls all four winners. Only Trend clears t > 3. That gap is the whole reason to demand the t-stat and not stop at profit factor.

So why keep the soft-t sleeves at all? Because they cover ground the Trend sleeve cannot. The VWAP Short sleeve, the one that reads a soft 2.56 on its own trades, is what carries the book through market falls: across the NQ declines of 10% or more from a record high, it made more than the whole book made there, while Trend and Opening Range Break lost money. That is why a soft t-stat can still earn its seat. How we test the book for overfitting is in is my backtest overfit, and each decline is in how the book did in each NQ decline.

So this check has a subtle rule. Demand the t-stat for whatever you are actually going to trade. If someone sells a single strategy, its solo t-stat is the one that counts. If they sell a book, the book number is fair, but a good one will still show you the parts underneath.

How do I tell if a strategy's edge held up over time?

Cut history into equal blocks and check profit factor in every one. Our book reads 1.05, 1.21, 1.37 and 1.73 across four four-year eras, so every block made money and the edge strengthened over time.

Ask whether the edge held up in each stretch of history, or whether one lucky run carries the whole record.

Profit factor is gross wins divided by gross losses. Above 1.0 means the period made money. A single aggregate profit factor can hide a record that made everything in one wild year and bled the rest of the time. The fix is to cut history into equal blocks and check every one.

Bar chart of profit factor by non-overlapping four-year era for the NQ book, with a red dashed break-even line at 1.0. Four green bars rising left to right: 2011 to 2014 profit factor 1.05 on 745 trades, 2015 to 2018 profit factor 1.21 on 909 trades, 2019 to 2022 profit factor 1.37 on 741 trades, 2023 to 2026 profit factor 1.73 on 744 trades. An amber trend line connects the bar tops to show the monotonic rise. Every era sits above the 1.0 break-even bar. Bar chart of profit factor by non-overlapping four-year era for the NQ book, with a red dashed break-even line at 1.0. Four green bars rising left to right: 2011 to 2014 profit factor 1.05 on 745 trades, 2015 to 2018 profit factor 1.21 on 909 trades, 2019 to 2022 profit factor 1.37 on 741 trades, 2023 to 2026 profit factor 1.73 on 744 trades. An amber trend line connects the bar tops to show the monotonic rise. Every era sits above the 1.0 break-even bar.
Profit factor by four-year block: 1.05, 1.21, 1.37, 1.73. Every era cleared break-even, and the edge got stronger over time, which is the opposite of what a curve fit does as it ages.

Split into four equal four-year eras, our book reads 1.05, 1.21, 1.37, and 1.73. Every block made money. The earliest is thin, barely above break-even on 745 trades, and we will say that plainly rather than round it up. But the trend runs the right way. A curve-fit strategy usually looks best on the data it was built on and decays as it ages. Ours strengthens. That is the pattern this check is built to find, and the thin early era is the honest cost of showing it.

What cost per contract should a trading track record include?

A real record names its cost assumption. Ours is about $4.10 round-turn, applied to every fill, so the $696,892 net is already post-cost.

Ask whether commissions and slippage were subtracted, and get the per-contract figure.

This is where a lot of records quietly cheat. A backtest with zero costs can turn a losing strategy into a winning one, especially a high-frequency one. The tell is a record that will not name its cost assumption. A real one gives you a number.

Ours is about $4.10 round-turn, applied to every fill. That is the configured, observed cost from our execution calibration, and our net P&L is post-cost by construction, so the $696,892 you see already has it taken out. You can check it trade by trade: each of the 3,139 trades nets out to its price move at $20 a point for each contract, minus $4.10 for each contract, and commission comes to $22,238 across the whole record. The point of the check is not the exact number. It is that a serious record can tell you the number at all, and a scalping strategy that skips it is selling you a fantasy.

What should a drawdown percentage be measured against?

Always ask for the denominator, because the same dollar loss can be very different percentages. Our worst $25,342 dip is 3.1% of the $804,557 peak, 3.6% of net, or 25.3% of a $100k account.

Ask what the drawdown percentage is divided by. The same dollar loss can be three very different percentages.

This is the check most people skip, and it is where a lot of marketing hides. A drawdown is the worst peak-to-trough drop in equity. The dollar figure is objective. The percentage is a choice, and the choice can flatter or scare depending on what sits in the denominator.

Horizontal bar chart showing the identical 25,342 dollar worst-dollar drawdown expressed as three different percentages against three different denominators. Green bar, 3.6 percent of final net profit of 696,892 dollars, labeled flattering. Blue bar, 3.1 percent of the 804,557 dollar equity peak it fell from. Red bar, 25.3 percent of a 100,000 dollar starting account, the harshest read. A note warns that a track record quoting a drawdown percent without its denominator is hiding the ball. Horizontal bar chart showing the identical 25,342 dollar worst-dollar drawdown expressed as three different percentages against three different denominators. Green bar, 3.6 percent of final net profit of 696,892 dollars, labeled flattering. Blue bar, 3.1 percent of the 804,557 dollar equity peak it fell from. Red bar, 25.3 percent of a 100,000 dollar starting account, the harshest read. A note warns that a track record quoting a drawdown percent without its denominator is hiding the ball.
The same $25,342 worst-dollar dip is 3.6% of net, 3.1% of the $804,557 peak it fell from, or 25.3% of a $100k account. One dollar figure, three honest percentages. Always ask which one you are being shown.

Our worst dollar drawdown was $25,342. It fell from a peak of $804,557, so against that peak it is 3.1%; against the final $697k net it is a small 3.6%; against a $100k starting account it is 25.3%. Same dollar figure, three honest percentages, and a marketer would quote the 3.1%. It began on August 4, 2026 and was still open when the data ends. Keep the denominators straight and do not let a big-account percentage get quoted as a small-account one. Our largest percentage drawdown is a separate, earlier episode: 20.4%, the figure on our tear sheet, a $20,432 dip below the $100,000 starting account that bottomed in September 2013, measured against the account's peak equity at the time. The check is simple: if a record shows a drawdown percentage and will not say the denominator, treat the number as marketing until proven otherwise.

Is a backtest's max drawdown the worst I should expect?

No. Our $25,342 backtested drawdown sat at the 5th percentile of 10,000 reshuffles, so about 95 of every 100 drew down worse. The median reshuffle is $34,223, and 1 in 20 went past $51,684.

Ask for the drawdown a reshuffle of the trades implies, not the single number the backtest happened to produce.

The realized drawdown is one roll of the dice. It depends heavily on the order the trades arrived in. Cluster the losers and the dip is deep. Spread them out and it is shallow. The backtest shows exactly one order, the one history dealt. A fair audit asks what a normal order would have produced.

We took our backtested trades and reshuffled the sequence 10,000 times. Our backtested $25,342 drawdown, 25.3% of a $100k account, landed at the 5th percentile of that distribution, meaning about 95 of every 100 reshuffles drew down worse than the backtest did. The median reshuffled drawdown is deeper than what the backtest produced, $34,223, and 1 in 20 reshuffles went past $51,684. So plan for more than our worst: the planning drawdown we use is the median, rounded up to $35,000, and it is not a ceiling.

We ran a second, stricter reshuffle to check that number. The simple shuffle treats every trade as independent, which can overstate the tail because it breaks up the losing streaks that actually drive drawdowns. So we also used a block reshuffle that keeps each losing streak intact. It comes out kinder to us at every percentile. We are not quoting its exact figures, because a fresh run moves them by more than we are willing to call reproducible, and a number we cannot reproduce has no business in an audit article. The direction is the finding: the simple shuffle is the conservative one, so it is the band we size against.

Confidence-band chart of max drawdown against a 100,000 dollar account. A blue marker sits at the backtested 25,342 dollars, 25.3 percent of the account, labeled backtested. A red simple reshuffle band runs from the 34,223 dollar median to the 51,684 dollar 1-in-20 case, with the backtested marker below it at the 5th percentile. A green streak-preserving block-reshuffle band sits lower than the red one, drawn unlabelled because its exact values are not reproducible to our standard. Confidence-band chart of max drawdown against a 100,000 dollar account. A blue marker sits at the backtested 25,342 dollars, 25.3 percent of the account, labeled backtested. A red simple reshuffle band runs from the 34,223 dollar median to the 51,684 dollar 1-in-20 case, with the backtested marker below it at the 5th percentile. A green streak-preserving block-reshuffle band sits lower than the red one, drawn unlabelled because its exact values are not reproducible to our standard.
The backtested $25,342, 25.3% of a $100k account, sits below the $34,223 median reshuffle, and 1 in 20 reshuffles went past $51,684. Plan for more than the single backtested ordering.

A record that quotes only its realized drawdown is showing you one draw dressed as a limit. The full method, and why the edge is still real either way, is in expect a worse drawdown than your backtest.

Should a good trading track record show losing years?

Yes. Real edges lose regularly. Our book had three negative years, 2011, 2012 and 2013, and 67 negative months out of 181, with a worst streak of 11 losses in a row for -$11,955.

Ask to see the down years and down months. A record with no red in it is either very short or not telling you everything.

Real edges lose regularly. Ours wins 45.3% of trades, so it loses more often than it wins, and the profit comes from wins being bigger than losses. Over 15 years the book had three negative years, 2011 at -$2,112, 2012 at -$1,708, and 2013 at -$3,985, and 67 negative months out of 181. The worst losing streak was 11 trades in a row for -$11,955, from October 14 to October 27, 2020, which is 1.72% of total net. None of that is hidden, because a record that only shows green is hiding the half of the data that tells you whether you can actually sit through it.

Bar chart of net profit by year for the NQ book from 2011 to 2026, with a zero line. Three red bars below zero mark the only losing years: 2011 at minus 2,112 dollars, 2012 at minus 1,708 dollars and 2013 at minus 3,985 dollars, all labeled. Every other year is a green bar above zero, growing from several thousand dollars in the early years to a high of 135,050 dollars in 2024. A red callout box notes the worst losing streak of 11 trades in a row for minus 11,955 dollars, October 14 to October 27, 2020, equal to 1.72 percent of net. Bar chart of net profit by year for the NQ book from 2011 to 2026, with a zero line. Three red bars below zero mark the only losing years: 2011 at minus 2,112 dollars, 2012 at minus 1,708 dollars and 2013 at minus 3,985 dollars, all labeled. Every other year is a green bar above zero, growing from several thousand dollars in the early years to a high of 135,050 dollars in 2024. A red callout box notes the worst losing streak of 11 trades in a row for minus 11,955 dollars, October 14 to October 27, 2020, equal to 1.72 percent of net.
Net by year, 2011 to 2026. Only 2011 (-$2,112), 2012 (-$1,708) and 2013 (-$3,985) were red, and we show them. The worst run inside any year was 11 straight losers for -$11,955, which is 1.72% of net.

The eighth thing to demand: a disclosure the checks cannot catch

There is one more thing to ask for that no scorecard captures, and it is not a grade. It is a disclosure. Ask whether the numbers can move when the underlying data gets re-based.

Continuous futures contracts get stitched together across rolls, and each roll re-bases the price series a little. So the exact same unchanged strategy can print slightly different totals on different runs, with nothing changed in the strategy itself. Our published book is $696,892 over 3,139 trades, exported on November 6, 2026. The data underneath can move, so every figure on this page carries the date of the export it came from.

This is why a screenshot of a number is weaker evidence than a dated per-trade export. A screenshot cannot show data re-basing; a process you can run again can. This is the same line that separates a real signal record from a highlight reel, which we lay out in are futures trading signals worth it. A track record that never mentions this, or that treats one screenshot as gospel, has not told you how its own numbers behave.

The one honest limit

Three of our four strategies do not clear the t-over-3 bar on their own trades. Overnight reads 2.93, VWAP Short 2.56 and Opening Range Break 2.09. We keep all three for the jobs they do inside the book, covering conditions the other sleeves sit out, not because any one of them is a bulletproof solo edge. An auditor applying check 2 strictly to each piece would flag three of ours, and they would be right to. We would rather you know that than discover it later.

Copy this checklist and run it on anyone

Paste these seven lines into your notes and point them at the next track record someone shows you, ours included.

  1. 200+ trades over 3+ years?
  2. t-stat over 3 for the thing you will actually trade?
  3. Profit factor above 1 in each 4-year era, not just the total?
  4. Costs subtracted, and can they name the per-contract number?
  5. Drawdown percentage, and what is the denominator?
  6. Planning drawdown quoted from a reshuffle, not the realized best case?
  7. Losing years and months shown, not hidden?

The honest providers answer with numbers. The rest change the subject to testimonials and one good week.

How we measured this

Instrument: CME Nasdaq-100 E-mini (NQ), continuous contract on TradingView, $100,000 starting capital, no compounding, 1 or 2 NQ minis a trade (2 when daily ranges are small, 1 when they are large; at today's index level every trade since July 2024 has been 1 contract). Window: 2011 to 2026 (data as of November 6, 2026). Fills modeled with TradingView's bar-magnifier; commissions and slippage in every trade, with commission at about $4.10 round-turn. The book is four sub-strategies run as one single-position portfolio.

Every book-level figure reconciles on the raw net-PnL column three independent ways: exit rows, entry rows, and the cumulative column all agree on 3,139 trades and $696,892. The per-era profit factors, the negative-year and negative-month counts, the worst streak, and the drawdown come from the same export. The reshuffled-drawdown figures come from a 10,000-iteration simple trade shuffle (median $34,223, 1 in 20 above $51,684); a streak-preserving block bootstrap runs below it at every percentile and is reported qualitatively only.

The limits, plainly. These are hypothetical backtest results, not live fills, so they inherit the engine's assumptions. The two reshuffles in check 6 disagree because the simple shuffle breaks up losing streaks while the block method keeps them, so the honest planning drawdown sits at or below the simple shuffle's band rather than above it. And the whole record leans on a recent, trend-friendly regime for the Nasdaq. The real risk to a book like ours is not a fake edge; it is a change in market character the past did not contain.

What to do with this

Run the seven checks on us before you trust anything else on this site. The full stat block behind every number here is on the tear sheet, and the four strategies, each with its own equity curve, are on our strategies page. If our book clears your version of these seven checks, the same systems send their entries as real-time signals, and plans are on the pricing page. We would rather you grade the receipts first than take our word for it.


We trade this book live and sell access to the signals, so judge the data accordingly. This article is educational and is not investment advice. Futures trading involves substantial risk of loss and is not suitable for every investor.

Hypothetical performance disclaimer (CFTC Rule 4.41): hypothetical or simulated performance results have certain limitations. Unlike an actual performance record, simulated results do not represent actual trading. Also, since the trades have not been executed, the results may have under- or over-compensated for the impact, if any, of certain market factors, such as lack of liquidity. Simulated trading programs in general are also subject to the fact that they are designed with the benefit of hindsight. No representation is being made that any account will or is likely to achieve profit or losses similar to those shown. Past performance does not indicate future results.

See what STS NQ futures signals are and how they reach you and our forward record, every live signal we have sent, dated and public.

Hypothetical performance disclosure (CFTC Rule 4.41). These results are based on simulated or hypothetical performance results that have certain inherent limitations. Unlike the results shown in an actual performance record, these results do not represent actual trading. Also, because these trades have not actually been executed, these results may have under-or over-compensated for the impact, if any, of certain market factors, such as lack of liquidity. Simulated or hypothetical trading programs in general are also subject to the fact that they are designed with the benefit of hindsight. No representation is being made that any account will or is likely to achieve profits or losses similar to these being shown.

Past performance is not indicative of future results. Trading futures involves substantial risk of loss.