Research

How we research

How we prove an edge

Before we trust a strategy — one we’re checking or one of our own — it has to clear the same six tests. Each one is built to strip out luck, cost, and hindsight, so what’s left is either a real edge or nothing.

Code it into exact rulesNo discretion, no exceptions. A rule a computer can’t execute identically every time isn’t a strategy we can test.
Run it on years of real tick dataTick-by-tick history across trends, chop, crashes, and quiet stretches — the full range of conditions, not a hand-picked window.
Charge real costs on every tradeCommission plus realistic slippage on entry and exit, applied to every fill. A cost-free backtest measures nothing that can be traded.
Measure it against luckA raw win rate means little on its own. We compare the strategy to random entries through the same stops and targets, and keep only the excess over chance.
Test it out-of-sample — onceBuild and tune on older history, hold back a recent stretch the strategy never saw, and run it there exactly once. Passing is evidence; failing means the result was fit to the past.
Account for the search, then publish it allThe more ideas you try, the more likely one looks good by chance — so we deflate for the number of attempts. Then we publish the hypothesis, the method, and the verdict, win or lose.
Anything that clears fewer than all six, we treat as luck until proven otherwise.

Below, that same standard is applied to strategies retail traders widely believe in. Each study publishes the full hypothesis, the method, and the verdict — win or lose.

The studies

Popular strategies, put to the test

Sep 2026

TJR’s Model

A widely-followed ICT intraday model, marketed at a ~64% win rate. We coded its exact rules and tested 22 reasonable readings of them on 7 years of NQ/ES tick data, after costs, in and out of sample.

His own rules: PF 1.06, Sharpe 0.16 · indistinguishable from random
Read the study →
Sep 2026

The VWAP “Holy Grail”

An academic paper calls VWAP trend trading the holy grail — 671% on QQQ at a 2.1 Sharpe. We reproduced it exactly on six years of 1-minute data, then charged the one cost it left out: the spread.

Reproduced at zero slippage · After the spread no edge
Read the study →
Sep 2026

Fair Value Gaps

The most-taught idea in smart money concepts. We found 5.6 million of them across seven futures markets and compared every one to the same size move with no gap.

5.6 million gaps · Hold about 1 point better than a matched move — too small to trade
Read the study →
Sep 2026

Inverse Fair Value Gaps

When price closes through a fair value gap, traders take it as confirmation the direction has flipped. We found 4.7 million of them across seven futures markets and compared every one to the same size move with no gap.

4.7 million gaps · Continue no more often than a matched move — no edge on any timeframe
Read the study →
Sep 2026

Liquidity Sweeps

Price takes out a key high or low, closes back inside, and is supposed to reverse. We found 37,162 sweeps across seven futures markets and compared every one to a sweep of a matched fake level.

37,162 sweeps · Reverse 49.9% vs 50.0% on a fake level — no edge
Read the study →
Coming soon

ICT Silver Bullet

The 10–11am ET "silver bullet" window is everywhere on YouTube. We're running it through the same test. Verdict published when it's done.

Coming soon
Reference

Key Terms, Explained

The metrics and concepts we publish, in plain English. If you followed a link to get here, your term is right below.

Edge
A real, repeatable advantage: a pattern that makes money across a large sample of trades, after costs, at a rate luck alone can’t explain. Most strategies sold online have none — they may win for a while by chance, but over enough trades they break even or lose.
Win rate
The percentage of trades that close in profit. On its own it says little: a high win rate paired with small wins and large losses still loses money. It only matters alongside the size of the wins versus the losses.
Profit factor
Gross profit divided by gross loss — for every $1 the strategy loses, how many dollars it makes back. Above 1.0 is profitable; 1.31 means $1.31 earned for every $1 given back.
Sharpe ratio
Return relative to risk — how much reward a strategy produces per unit of volatility. Higher is better: above 2 is excellent, above 3 is rare even among top funds. It rewards steady returns and penalizes wild swings.
Costs (commission & slippage)
The two frictions charged on every real trade. Commission is the fixed fee per contract, paid on entry and exit. Slippage is the gap between the price you wanted and the price you got, worst around news and the open. Every figure we publish is calculated after both.
Backtest (and multi-year backtest)
Running a strategy against years of historical data to see how it would have performed across many market conditions. Backtested results are hypothetical, not live trading, and are only meaningful once realistic costs are included.
Walk-forward testing
A stricter backtest: build and tune on one stretch of history, test on the next stretch it never saw, then repeat forward through time. It mimics trading the strategy live rather than fitting it to the whole past at once.
Out-of-sample
A block of history set aside during development and never used to build the strategy, then used once to test the finished version. Holding up on data it has never seen is genuine evidence of an edge, rather than a fit to the past.
Noise
Random, meaningless movement in the data. Markets throw up patterns purely by chance that fit the past perfectly and fall apart on new data. Separating a real, repeatable edge from noise is the whole point of honest testing.

The strategies that clear the same bar.

We hold our own strategies to the identical test. Their full track records — after costs, out-of-sample — are on the Performance page.

See the Performance View pricing