Anti-Martingale: Add to Winners
The anti-martingale increases position size as a trade moves in your favour and never as it moves against you. It is the inverse of the martingale, which doubles the stake after each loss. The stop must rise with every addition, so the whole position never risks more than the plan allowed.
Adding to a winner is the only form of averaging that survives arithmetic. It is also the hardest thing here to actually do.
How it works
The rule is one sentence: add size as a trade moves in your favour, never as it moves against you. Each later tranche needs a defined further move first.
The martingale is the opposite scheme — double the stake after every loss. Turn it the right way up and you get this. Grid trading is its chart-shaped cousin and fails the same way.
The inversion is sound because of where the risk sits. A loser’s risk is uncapped by the size you keep adding; a winner’s is capped by a trailing stop.
By the first addition you are risking the market’s money, not the original stake — provided the stop has moved. The win rate falls and the expectancy does not.
It loses often and small, and wins rarely and big. Which is arithmetically sound and psychologically the hardest shape there is to sit through, because most outcomes are unpleasant.
Where the additions go
Every addition needs a rule written before the trade, not a feeling during it. A new swing high, a measured advance, a multiple of the average true range (ATR) — “it looks strong” is trading psychology deciding for you.
And adding raises risk on a position that is already working. So the stop rises with each tranche, keeping the whole position inside what risk per trade allowed at the start.
Each addition pays the spread all over again. A plan with many small tranches is a plan with a large cost bill, whether or not the trend continues.
A thinning market is the wrong place to add. Falling volume into an advance means fewer participants behind the move — exactly when a larger position is least welcome.
In practice
This needs a long trend, and long trends are the minority state. On the history measured below, direction runs average 2.01 bars — most of the time there is nothing to add to. It suits trend following and position trading.
An opening gap can take back every addition at once. A stop names a level, not a price, and a market that opens beyond it fills the enlarged position somewhere worse.
So the stop moves up with each add, or this is not the anti-martingale. Trailed by a multiple of the 14-bar average true range, the median position here survived 3 bars at one average range, 10 at two, 22 at three and 32 at four, across 562 trials each. Between 91% and 100% were eventually stopped out.
Every round trip costs 0.0098 price units — 2% of a median bar’s range, and 45% of the smallest bar. Multiply by the tranches you intend for the real fee.
Sizing the additions
There are three ways to size the tranches and only one is easy to defend. Equal-size additions are the simplest to plan and the most expensive, because each one drags the average entry higher, so a routine pullback can take a comfortable position back to break-even.
Increasing size is worse again. The largest tranche is bought closest to the eventual high, which is the least attractive price in the sequence.
Decreasing size — halving each time, or something near it — is the more defensible default. It keeps the average entry near the original, and leaves the bulk of the position bought at the price you actually liked.
The trade-off is real and worth stating. A smaller final position, and a smaller win when the trend does run — which, on a series where runs are this short, is a price worth paying.
What the anti-martingale is not
It is not a way to be right more often — the win rate usually falls.
It is not risk management — it is a sizing rule that needs one underneath.
It is not a repair for a losing position — that is the martingale, and it is what this avoids.
And it is not free — every tranche pays the spread again and lifts the average entry.
When it fails
In a trading range it bleeds slowly and correctly. Each false start triggers an addition, the move reverses, and the enlarged position is stopped for more than the original would have lost. Nothing malfunctioned — the method paid for information.
The second failure is the average-entry surprise. Traders add, watch a routine pullback erase the open profit, and conclude the trade went wrong — when what moved was their own break-even.
A third is adding without moving the stop loss. That is not this method at all — it is a bigger position wearing the vocabulary of a smaller one, and the drawdown scales with the size, not the plan.
A fourth is the cost stack. Many tranches means many spreads, quietly consuming the edge the rare large winner was supposed to provide.
And a fifth is misreading probability. A trend is not more likely to continue because it has already run. The method never claims otherwise — only that losses are capped and winners are not.
The original data
One video in 31,760 titles is called “adding to winners”, and it has 359,180 views. One video, one
channel, and among the most-watched things measured anywhere on this site. “Pyramiding” appears once,
at 50,845 views. “Anti-martingale” appears zero times, as do “risk of ruin” and “expectancy”, while
“martingale” appears once at 23,800 views against 211 videos on “win rate” at a median of 11,527.
Those counts are in research/broker-coverage.json: the method is known by a name nobody searches and
searched for under a name nobody uses.
And then the limit. On this site’s shared 576-bar history in research/series-measurements.json,
produced by site/measure_series.py, direction runs average 2.01 bars with a longest of 11 across 286
runs, and the ten-bar efficiency ratio has a median of 0.34 with 30% of bars above 0.5. Most of the
time there is nothing to add to — the honest limit of the method, not an argument against it. Write
down where each addition happens and where the stop goes after it before you enter; if raising the
stop with the add is not in the plan, this is not the anti-martingale, it is just a bigger position.
Related
Risk per trade sets the budget every addition has to stay inside. Trend following is the approach this rule was built for, because it produces the moves long enough to add into. And trailing stop caps the enlarged position, which is what makes this a method rather than a larger bet.
Adding to a winner feels wrong in a way that adding to a loser never does, which tells you something about how badly wired we are for this. When a position is finally working, every instinct says protect it, bank it, do not touch it - and the plan says put more on. I have talked myself out of good additions far more often than I have talked myself into bad ones. The only thing that ever fixed it was writing the add level down before I entered, so the decision was already made by someone calmer.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.