Issue 3 · Tuesday, November 3, 2026 · 5-minute read
Disclosure: I don't trade the 3 Bar Play, and I have no connection to Jared Wesley or Live Traders. No one paid for or reviewed this test.
Good morning. Last week, the 3 Bar Play failed our test: 168 sealed trades that broke even after costs, with a p-value of 0.37 against random entries. Today: how the same test shows traders getting fooled.
WEEKLY QUIZ Intermediate
Last week's answer: B, 33%
At 2:1, one win pays for two losses: 1 ÷ (1 + 2) = 33% breaks even before costs. After costs, it needed 35.7%.
In 2023 to 2025, TSLA averaged +0.64R per trade, second best of our 22 stocks. If none of the 22 had any edge, how often would the best one average +0.64R or more?
Read today's results, then vote. The answer arrives next Tuesday.
In 2023 to 2025, TSLA averaged +0.64R per trade, second best of our 22 stocks. If none of the 22 had any edge, how often would the best one average +0.64R or more?
LUCK OR EDGE? · THE 3 BAR PLAY · PART 2 OF 2
How a strategy with no edge made nearly 30R in five months
Picture a trader who started following our 3 Bar Play rules on January 2, 2026, on all 22 stocks. By May 18, they were up 29.4R: about $2,935 at $100 of risk per trade. 79 trades, a 48% win rate, and every reason to believe.
By September 30, it was gone. The year finished almost exactly at zero: −0.04R.
Same rules, two different stories
| Jan 2 to May 18 | May 19 to Sep 30 |
|---|---|---|
Trades | 79 | 89 |
Win rate | 48% | 25% |
Average per trade, after costs | +0.37R | −0.33R |
Total | +29.4R | −29.4R |
The 2026 result after costs, trade by trade (teal), against 40 simulated traders with no edge at all (grey), picked at random from 10,000. Each simulated trader takes 168 trades drawn from the real 3 Bar Play results, shifted so the average is exactly zero.
Why a strategy with no edge can make money for months
Wins are twice the size of losses. Each trade adds about +1.9R or takes away 1.0R, so the running total swings hard, even with no edge.
Good runs are common. Of our 10,000 simulated traders with no edge, 11% got as high as our May peak of +29.4R. In p-value terms, the run-up has p = 0.11: luck does it about 1 time in 9. Half the simulated traders also went through a drawdown of 20R or more.
Wins cluster. In February, 59% of trades won. In September, 16% did. Nothing about the rules changed.
The trader who stopped in May would tell you the 3 Bar Play works. The one who started in May would say it doesn't. Both would be describing luck, and the results shared online tend to come from the good runs.
Why we needed 22 stocks
The 3 Bar Play is rare: about 9 trades a year per stock. On a single stock, 100 sealed trades means more than 10 years of waiting. With 22 stocks, nine months of 2026 gave us 168.
Many stocks also protect you from the one that looks like a winner by luck. In 2023 to 2025, TSLA made +0.64R per trade over 43 trades, profitable three years in a row. In 2026: −0.13R per trade.
When we simulated 22 stocks with no edge, the best one typically averaged about +0.56R per trade. The best of many always looks good: judge it against the best that luck produces, not against zero.
Average result per trade after costs, by stock. Grey: 2023 to 2025, with the range it would likely fall in if the test were repeated. Teal: 2026, from 2 to 15 trades per stock, too few for a meaningful range. NVDA had no 2026 trades.
The ranking didn't carry over, either. The five best stocks of 2023 to 2025 lost 0.38R per trade in 2026. The five worst made +0.50R.
Longs vs shorts: the sides swapped
| 2023 to 2025 | 2026 | All four years |
|---|---|---|---|
Longs | +0.10R | −0.18R | +0.03R |
Shorts | −0.03R | +0.26R | +0.03R |
Average result per trade after costs.
Trading only longs after 2023 to 2025 would have lost 18.0R in 2026, while the dropped shorts made 18.0R. Over all four years, both sides ended at +0.03R per trade. The first half hour did the same: best in 2023 to 2025, negative in 2026. Every filter found by looking at the results is one more chance to fit noise.
The caveats
Our rules are one translation of the pattern. "Big bar" and "tight pause" had to become numbers, picked before any result. Two variants and four chart sizes were tested too. None passed.
Large stocks only. Small caps with news and pre-market trading weren't tested.
The test can miss a small edge. 168 trades would catch an edge of +0.10R per trade only about 1 time in 3. A large one would most likely have shown up.
APPLY IT THIS WEEK
Count trades, not months. Five great months can be 79 trades of luck. Ask how often a no-edge strategy with the same win rate and payoff would do as well: that's your p-value.
Test across many instruments. A setup that fires 9 times a year needs 20 or more of them.
Treat filters found in the results as new ideas. Longs only, best stocks, best hour: each needs its own test on new data.
TERM OF THE WEEK
Regression to the mean
When results are partly luck, the extremes tend to move back toward average next time, like TSLA's +0.64R falling to −0.13R. The best result on a list is usually a worse bet than it looks.
NEXT WEEK
Live trading is too slow to tell luck from edge
Why judging a strategy on live results can take years, and how a backtest gets you there in an afternoon.
Your turn
Hit reply: have you had a great run that disappeared? How many trades did it last?
P.S. Know a trader about to go live with a strategy? Forward this. It's free for them too.
LuckOrEdge is educational content only, not investment advice. This continues the test of the mechanical rules from Issue 2, not any person's full trading method. Interactive Brokers 1-minute data, $0.99 per order, 1¢ per share of slippage, $100 of risk per trade. Simulations resample the real trades and are illustrations, not forecasts. Past performance does not guarantee future results.