Am I Overtrading? What Your Own Trade Data Shows

Overtrading is usually described as a discipline problem, and the advice that comes with it tends to arrive as a rule: a maximum number of trades, a stop after so many losses. This post does not offer a rule. It describes what overtrading tends to look like in a trader's own records, and at the arithmetic that makes extra trades more expensive than they feel.

The short version is that overtrading is not really about the number of trades. A trader who takes twenty trades a day on a setup with a real edge, at costs that edge can carry, is not overtrading. A trader who takes three, where the third has no edge at all, may be. The question is not how many, but what the marginal trade earns after it has paid its way.

Results by trade number

The most direct check is simple to describe. Number each trade by its position within the day: the first trade of the session is trade 1, the next is trade 2, and so on. Then group the whole history by that number and compare the averages.

The table below is a hypothetical illustration, not real data. It describes 100 trading days. Every day has a first trade, fewer days have a second, and fewer again have a fifth. Costs, commission plus slippage, are assumed at a flat £6 per round trip.

The book as a whole is profitable: 345 trades, £2,975 net. Read only at that level, nothing looks wrong. Split by trade number, the picture changes. The first two trades of the day carry £3,370, more than the entire result. The third roughly breaks even. Trades four onward lose £535 between them.

Where the losses on extra trades come from

Look at the gross column for those later trades. Across the 85 trades numbered four and above, the gross result is minus £25: 45 trades at plus £3 and 40 at minus £4. At the price level they are essentially break-even. They are not bad trades in the sense of being wrong about direction. They are trades with no measurable edge.

A trade with no edge is not free. Those 85 trades paid £510 in costs, which is almost the entire £535 they took from the book. This is the part of overtrading that is easiest to miss: a trade with zero edge is not neutral, it is an expected loss of exactly what it costs to trade. The 4th trade makes the same point more sharply. It still has a small positive gross edge, £3 per trade, but that sits below the £6 it costs to take, so it loses money while being right on average. The break-even cost method frames this as gross edge per trade against cost per trade, and the same comparison works row by row.

The flat £6 is also a simplification that probably flatters the later trades. If extra trades are taken faster, with market orders, in moving conditions, their slippage can be worse than the slippage on a planned first entry. Measured cost per trade, split by trade number, is often more informative than an assumed one.

Is the decline real?

A falling column is not yet a finding. Two things deserve a check before reading anything into it.

The first is noise. Suppose the standard deviation of net results from one trade to the next is around £60. The average of 40 trades then has a standard error of about £9.50, so a net average of minus £10 is roughly one standard error below zero. On its own, that row is hard to tell apart from break-even. The first row is a different matter: 100 trades averaging plus £22, with a standard error of £6, is about 3.7 standard errors above zero. The steady pattern across rows says more than any single row does, and a history several times this size would say more again. How many trades it takes to separate a real difference from a lucky one is its own subject.

The second is confounding. Trade number is tangled up with other things. A fifth trade is usually a later trade, so the row partly measures time of day. It also exists only on certain days: busy days, which may be choppy days, or losing days where the extra trades are attempts to recover, the pattern described in what revenge trading looks like in your data. One useful comparison holds the days fixed. On the days that had five or more trades, how did the first trade do? If it was as good as first trades elsewhere, the decline belongs to the later trades. If it was also weak, the problem may be the conditions on those days rather than the count.

More opportunities or more noise

None of this means more trades are worse. Sometimes a busy day is busy because the market offered more of what the setup needs. The data can usually tell the two cases apart.

When extra trades are extra opportunities, they tend to look like the first ones: the same setup, similar conditions, and a gross average per trade that holds roughly level as the count rises. Net results per day climb with the number of trades taken.

When extra trades are noise, they tend to drift away from the first ones: different setups, looser entries, or the same setup in conditions where it has no record. The gross average decays towards zero as the count rises while costs stay constant or grow, and net results per day flatten or fall on the busier days.

NoxarQuant's tags give one way to see that drift. Every trade is tagged with its setup, session, volatility bucket, RSI bucket and trend, and trades sharing the same tags form a cluster. Position in the day is not one of those tags, but session catches part of it, and the clusters that later trades fall into are worth a look. If first trades land mostly in clusters classified strong or promising, while fourth and fifth trades scatter across clusters too thin for a verdict (the classifier gives none under 12 trades) or into clusters that lost money, that is the noise pattern in a different format.

What the data cannot tell you

A table like this describes where results came from. It does not say what a trader ought to do about it, and it does not supply a number. A cap placed where this table turns negative would have improved the book in hindsight, but hindsight is exactly how the table was built. A rule chosen from the same data that suggested it will look better on that data than it does on the next hundred days, which is the general problem of overfitting. The useful output is a description: which trades carry the edge, which trades only carry the cost, and whether the difference between them is larger than noise.

Whether your own later trades look like your first ones or like the bottom rows of that table is a question your trade history can already answer, and it is the kind of measurement our case study is built around.