Ten good trades prove almost nothing
You want a number, and anyone who gives you a confident one is guessing. What can be shown: why small samples lie, how streaks run, and how to test a setup in blocks.
A coin flip can show you a 70 percent win rate
Flip a fair coin ten times. Seven or more heads turns up about one time in six. That is 176 of the 1,024 possible sequences.
So a method with no edge prints a 70 percent win rate over ten trades, one block in six. Your new setup went 7 and 3 last week. That result is completely compatible with the setup being worthless, and completely compatible with it being excellent. Ten trades cannot tell the two apart.
This cuts the other way too, which is the expensive half. A genuinely good setup can go 3 and 7 out of the gate and get abandoned. Traders drop working methods during ordinary bad runs. Then they buy the next one at the top of its lucky streak.
Losing streaks are shorter than they feel and more common than you think
Five losses in a row at a 50 percent win rate is 0.5 to the fifth power. That comes to 1 in 32. At a 40 percent win rate it is 0.6 to the fifth, roughly 1 in 13.
Those numbers describe any single run of five trades. You do not take one run of five, you take dozens. Over a hundred trades a five-loss streak stops being a surprise and becomes something to plan the week around.
That matters for two reasons. The streak is what makes you quit a fine strategy. It is also what makes you double size to get it back. Knowing the arithmetic in advance turns a scary run into an expected one.
Write your own version of these numbers using your actual win rate. Do it before you need it.
Test in blocks you commit to before you start
Pick one setup. Decide the number of trades in the block, decide the size, and write both down before the first entry.
Many traders work in blocks of 30 to 50 trades per setup. Treat that as a unit of measurement rather than a proof. The honest number depends on the size of your edge. A setup averaging plus 3R declares itself faster than one grinding out plus 0.4R.
Inside the block, size stays flat and the rules stay frozen. Change one variable mid-block and you have two half-samples of two different strategies.
At the end, record three things. Expectancy per trade in R, the worst streak, and the deepest drawdown in R. Then run a second block. Two blocks that agree tell you more than one block twice the size.
Judge each setup on its own record
Your account total is a blend. Blends hide the thing you are trying to see.
A typical account has one setup carrying everything and two others living off it. The overall equity curve looks flat, so you conclude your whole approach is broken. Split by setup tag and the picture usually resolves in a minute. Gap fades are paying and the afternoon breakouts are giving it back.
Tag on three axes and you can answer almost any question later. The setup name, the time of day, and the market condition you thought you were in.
One warning about this. Slice thin enough and every trader finds a beautiful subset, because with enough filters random data produces one. Decide your tags in advance and stick to them, instead of hunting for the cut that flatters you.
Set up a block test this week
Ten steps in all. Eight of them happen before your first trade of the block.
- Name the single setup being tested and write its entry rule in one sentence.
- Write the stop rule and the target rule in one sentence each.
- Fix the number of trades in the block and write it down.
- Fix the risk per trade in dollars and keep it flat for the whole block.
- Estimate your expected worst streak from your own win rate.
- Set a hard stop for the test itself, in R, at which the block ends early.
- Tag every trade in the block with the setup name and the time of day.
- Log realised R for each trade on the day it closes.
- At the end, compute expectancy in R, the worst streak and the deepest drawdown.
- Run a second block before you change anything or add size.
Frequently asked questions
Not on its own. Twenty trades is enough to cut your size and to check you followed the rules. Both are worth doing.
At a 40 percent win rate, five in a row comes up about one run of five in thirteen. Across a hundred trades you should plan on seeing it more than once.
Keep them in a separate column. Backtest fills skip slippage and skip the trades you talk yourself out of. Mixing them in flatters the live record.
Build the sample inside TraderLog
Trades import from Schwab and IBKR automatically, so the count grows without you typing. The stats page reports win rate, expectancy and profit factor across whatever you have so far. Results by symbol and by hold time show where the sample is still thin.
Every trading morning at 8:40 ET our model draws the SPY, QQQ and IWM zones it expects to matter, on a TradingView chart, before the open. Where price reaches one, it has turned 74 percent of the time. Free, no account. See today's map