How to know if you're overtrading: three tests

Every article on overtrading gives you the same list of symptoms: you trade when you are bored, you chase, you revenge trade. All true, and all useless at 14:40 on a Tuesday when you are looking at a fourth setup and it genuinely looks fine.

The question you actually need answered is narrower: for me, at my frequency, does the next trade still carry an edge? That is a number, and it is already sitting in your journal.

Overtrading has a definition you can test#

Overtrading is not "too many trades". Some profitable traders take thirty a week; some take three. The number by itself means nothing.

The useful definition is this: you are overtrading when your expectancy per trade falls as your trade count rises. Volume is only a problem when the marginal trade is worse than your average trade — when trade number five on a given day is not the same quality of bet as trade number one.

That flips the question from a feeling ("am I trading too much?") into a comparison you can run on your own history. You need three things logged for every trade: the date, the outcome in R, and the strategy or setup you were taking. If your journal has that, you have everything.

One condition, and it is the one that breaks most attempts at this: it only works if you logged the trades you are embarrassed about. A journal missing the impulsive Friday-afternoon scalps will tell you that you do not overtrade. That is not a result, it is a gap in the data.

Start with one number: expectancy per trade#

Expectancy is what an average trade is worth, expressed in R so that trade size does not distort it.

Expectancy = (Win% × average win in R) − (Loss% × average loss in R)

Take a sample of 168 trades as a worked example. 40% of them won. The average winner returned 1.9R, the average loser cost 1.0R:

(0.40 × 1.9) − (0.60 × 1.0)
= 0.76 − 0.60
= +0.16R per trade

Positive. Sixteen hundredths of your risk per trade, on average. Now hold that number, because the top-line figure is exactly where the problem hides.

If you are new to thinking in R, the R-multiple explainer covers the conversion from currency to R first — the rest of this post assumes it.

Test 1: sort your days by trade count#

Group every trading day by how many trades you took that day, then compute the R per trade inside each group. Illustrative numbers from a 60-day sample:

The aggregate is +0.11R per trade, and it is a lie of averages. The quiet days earn five times the aggregate. The busy days lose money outright — and they are not a rounding error, they are 55 of 168 trades, a third of everything you did.

That is the shape of overtrading when you finally see it in numbers. Not a catastrophe, just a steady leak that the profitable days keep paying for.

Two practical notes. Group in ranges, not exact counts — "exactly 7 trades" will have four days in it and tell you nothing. And use net R, not currency, so that a single oversized position does not decide the answer for you.

I ran this on my last 60 trading days — 135 trades, 17 March through 17 September. The aggregate is +0.05R per trade, and it is doing the same lying the table above does. I never took more than five in this window, so the exact counts had enough days in them and I left the ranges alone.

It does not cross zero once. One-trade days print. Two-trade days are a hole — and on 11 of those 12 days the first trade was already a loss, so the second look is probably revenge rather than a second setup. Three-trade days are fine. Four or more is where frequency itself goes negative. The surprise was not the busy days. I had been watching for "too many". The leak was the day I took exactly one more after a red and then stopped.

Test 2: find where the day turns#

The bucket table tells you that busy days are worse. It does not tell you which trade made them worse. For that, number the trades within each day — first trade of the day, second, third — and compute R per trade by position:

Same 168 trades, same +18.3R in total, cut a different way. The decay is monotonic, and it crosses zero at the third trade of the day.

This is the more actionable of the two tests, because it gives you a rule with a clear trigger rather than a vague resolution to trade less. It also survives the obvious objection to Test 1 — that busy days might simply be high-volatility days where more trades are legitimately available. If extra opportunity were the whole story, the fourth trade would earn roughly what the first one does. Here it does not.

Watch the sample size on the bottom row. Thirty-three trades is enough to notice a pattern and not enough to be certain of its size; treat it as a flag, not a verdict.

Test 3: check that it is frequency, not the setup#

Before you blame volume, rule out the simpler explanation: that your extra trades are a different, worse strategy wearing the same clothes.

Split the same trades by strategy tag and look at the count distribution. In most journals, one or two setups account for almost all of the late-in-the-day entries — the lower-conviction continuation trade, the second entry after a stop-out, the pair you only touch when nothing else is moving. If the negative R is concentrated there, you do not have a frequency problem. You have one setup that does not work, and it happens to be the one you reach for when you are bored.

The distinction matters because the fixes are opposite. A frequency problem is solved with a hard stop on trade count. A bad-setup problem is solved by removing that setup — and cutting your trade count instead would also cut the good trades, for no reason.

This is where the forex use-case breakdown and per-strategy tagging in your journal earn their keep: the analysis is only as good as the granularity you log.

I tagged the same 135 trades. They did not collapse into one late-day orphan. FCS is 46% of everything I took and 46% of the third-or-later trades — same clothes in the morning as in the afternoon. That is the tell. FCS is simply a losing setup: 62 trades, −10R, −0.16R per trade, and −0.50R even as the first trade of the day. Strip it out and the remaining 73 trades are +0.22R. The two-trade hole is mostly this: 12 of those 24 trades are FCS, all −1R.

SMC is the other half of the book, and it is the edge — 56 trades, +16.7R, +0.30R per trade. The first two SMC trades of the day print (+0.55R, +0.68R). The third and later do not. That is a frequency problem sitting on top of a good setup.

So I do not have one diagnosis. I have a setup I should stop taking, and a cap I should put on the setup that works. Cutting count without touching FCS would still leave the leak. Removing FCS without a cap would still spend the SMC edge on the third look.

What these tests do not tell you#

Intellectual honesty about the limits, because the tests are easy to over-read.

They do not prove causation. Days when you take six trades might be days when you started with a loss and were trying to recover — in which case the trade count is the symptom, not the disease, and a hard cap will not fix the state of mind that produced it.

They do not separate skill from variance on a small sample. Thirty-three trades at −0.30R is a plausible result for a strategy with a genuinely positive edge having a bad run. If you want to see the range of outcomes a given win rate and reward-to-risk can produce over a given number of trades, run it through the Monte Carlo simulator before you conclude anything. The same logic that makes a long losing streak ordinary applies here.

They do not include your costs, unless you logged them net. Costs scale with trade count, which makes them the most reliable penalty of high frequency. Work it out once: if 1R is $200 of risk and a round turn costs you $8 in spread and commission, that is 0.04R per trade. Across 168 trades, $1,344 — a third of the gross result in the table above. Your own figures will differ; compute them rather than assuming they are noise.

They do not tell you the counterfactual. "If I had stopped at three trades a day I would have made more" is not quite right — the trades you would have skipped are not necessarily the ones that lost. The tests tell you where the average is worse, not which specific trade to remove.

Turn the number into a rule you can follow#

A diagnosis you do not act on is entertainment. Three ways traders convert this into something that holds:

A hard count limit, set one below where your data crosses zero. In the worked example, the third trade is already break-even, so the rule is two and done. The point of setting it one trade early is that you make the decision when you are calm rather than when you are staring at a setup.

A checklist gate rather than a count. If the problem is quality rather than quantity, require the marginal trade to clear a higher bar — a written pre-trade checklist that the late-day entries must pass in full. Checklist templates in your journal make this a thirty-second step instead of an intention.

Log the trade you did not take. Record the setup you passed on and what it would have done. After a month you will know whether your limit is costing you anything, which is the only honest way to find out whether the rule is too tight.

Then re-run the two tables in a month. The tests are not a one-time diagnosis; they are how you check that the rule is doing what you set it up to do.

I do not run any of the three. No checklist. I do not log the trades I pass on. I do not have a number on the wall.

What the journal shows instead is that I stopped taking four. Since 1 June: 17 trading days, 23 trades. I have not taken a fourth once. A cap of three held. A cap of two I broke twice — 30 June (three trades, +3R) and 19 August (three trades, −3R). July I took one a day for six days and made +14R. That is the quiet-day bucket from Test 1, as a calendar month.

The week it did not hold is 19 August. Three trades, −3R. I had already seen the two-trade hole and I took the third anyway.

The rule the two tables actually ask for is tighter than three: two SMC and done, and no FCS. I have not written that down either. September already has two two-trade days, both −2R.

Overtrading is not a character flaw to be willpowered away. It is a measurable decay in the quality of your marginal trade, and it has a number attached. Pull 60 days of your own history, build the two tables, and you will know within twenty minutes whether the fourth trade of the day is paying for itself — which is more than the voice in your head has ever told you.