The Danger After Winning: Overconfidence, House Money and Sizing Up on a Streak

Many traders are careful after losses, yet the worst damage to an account often starts after a great day. Five wins in a row, a sense that you finally read the market, so you size up, take trades outside the plan and trade more often. This article looks at how often win streaks happen by pure chance, what research says about behaviour after gains, and then simulates the real numbers of sizing up on a streak.

It builds on trading biases and is the flip side of how long a losing streak is normal.

How often win streaks happen by chance

Say a system wins 50% and each trade is independent. The chance that one particular set of 5 trades all win is 0.5 to the power 5 = 3.1%. That sounds rare, but over 100 trades a streak can start on almost every trade. We computed the exact probabilities trade by trade (no random sampling):

• A system that wins 50% over 100 trades has an 81.0% chance of at least one streak of 5 or more wins
• The chance of 7 or more wins in a row is still 31.8%, and the average longest win streak is about 6.0 trades
• A system that wins only 40% still has a 45.9% chance of a 5+ win streak

A long win streak doesn’t prove you got better, just as a long losing streak doesn’t prove the system broke. Luck produces both on its own.

Why the brain reads a streak as skill

Self-attribution bias is the tendency to credit good outcomes to our own skill and blame bad ones on luck or outside forces. Miller and Ross (1975) reviewed the early evidence and found it clearest on the first side: taking credit for success. In trading it shows up as familiar lines. A winner: “read it perfectly”. A loser: “the news spiked me” or “the broker hunted my stop”.

Gervais and Odean (2001) built a model called “Learning to Be Overconfident”. If traders give themselves too much credit for winning trades, their confidence grows faster than their real skill, and it grows fastest early on, right after the first successes. The model doesn’t say early profits are all luck. It says the way we learn from profits makes overconfidence easy to pick up.

The catch is that peak confidence tends to arrive right after a win streak, and as the previous section showed, streaks happen by chance all the time.

The house money effect: profit doesn’t feel like your money

Thaler and Johnson (1990) ran experiments where people chose gambles after first winning or losing money. They found two patterns:

• House money effect. After a gain, people take more risk, like a gambler ahead at the casino who feels they are playing with “the house’s money”, so losing it hurts less than losing their own
• Break-even effect. After a loss, people also take more risk, when an option offers a chance to get back to even. That is the root of revenge trading

In a trading account, yesterday’s profit and your original deposit are one pool of money. Each dollar is worth the same. But the mind keeps separate accounts, so profit gets risked more freely. The body plays a part too: a run of wins can shift hormone levels and appetite for risk. See stress hormones and risk taking.

Overtrading: overconfidence leads to trading more

Barber and Odean studied real accounts of retail stock investors, not lab experiments:

• “Trading Is Hazardous to Your Wealth” (2000) used data on about 66,000 households from 1991 to 1996. The 20% who traded most earned about 11.4% a year net, while the market returned about 17.9%. Most of the gap came from commissions and costs paid on every trade, not from picking much worse stocks
• “Boys Will Be Boys” (2001) found that men traded about 45% more than women, and trading cut men’s net returns by more. The authors read this as consistent with overconfidence, since psychology research finds men tend to be more confident in financial matters
• Their 2002 paper followed investors who switched to online trading. This group had beaten the market before switching, but afterwards traded more often and more speculatively, and fell behind the market. It matches the theme here: earlier success leading to heavier trading

In Forex and gold the main costs are spread and commission. A trade opened on confidence but without an edge doesn’t average zero. It averages minus the cost. See spread cost per trade.

The hot hand: a debate over 40 years long

Gilovich, Vallone and Tversky (1985) looked at basketball shooting records and found that the chance of making a shot after a make was not meaningfully higher than after a miss. They concluded that the “hot hand” players and fans believe in was an illusion. The study became a classic example of seeing patterns in randomness.

Miller and Sanjurjo (2018) showed the measurement had a hidden bias. Flip a coin 4 times and ask, within that sequence, “after a head, what % of the next flips are heads?”

1. There are 16 possible sequences. Drop the ones with no head in the first 3 flips (TTTT and TTTH), leaving 14
2. Work out the % in each sequence. HTHT gives 0%, HHHH gives 100%
3. The average across the 14 sequences is 40.5%, not 50%, even though the coin is fair

The cause: a sequence with a long run of heads has many moments to count but still only counts as one value, while a sequence where a lone head is followed by a tail scores a full 0%. That drags the average down. When the authors corrected for this bias in the controlled shooting data from the 1985 study, the result showed evidence of a hot hand, reversing the original conclusion for that data. How big a hot hand is in sport is still being argued.

The same bias can show up in a trading journal. We simulated a pure coin-flip system that wins 50%, 100 trades per journal, 200,000 times, and in each journal measured “the win rate of trades that follow 3 wins in a row”. The average came out at about 46%. If you see this in your own journal, don’t conclude “after a streak, a loss is due”. That result can come purely from the way it’s counted.

Two lessons for traders. First, the feeling of being hot can’t be trusted. Second, the right question isn’t “am I hot right now?” but “are my system’s trade results really linked to each other?” That can only be answered with many trades, compared against the same trades shuffled into random order. See how many trades is enough and Monte Carlo testing.

Simulation: what sizing up on a streak really does

Take a hypothetical system with a small edge: it wins 50%, a winner makes 1.2R and a loser costs 1R (R is the money lost when a Stop loss is hit once). Set 1R = 100 dollars, 200 trades, repeated 20,000 times, with the same trade sequence used for every rule.

Average result per trade = 0.5 × 1.2 − 0.5 × 1 = 0.1R, or 10 dollars, so 200 trades should make about 2,000 dollars. The simulated fixed-size rule averaged 1,996 dollars, which matches.

Three rules compared:

• Fixed size. Risk 1R on every trade
• Double after 3 wins. After 3 or more wins in a row, the next trade risks 2R until a loss, then back to 1R
• Escalate on a streak. After 3 wins risk 2R, after 4 wins risk 4R, after 5 or more risk 8R. Back to 1R after any loss

After 200 trades (dollars)	Fixed size	Double after 3 wins	Escalate on a streak
Median result	+2,000	+2,180	+2,140
Mean result	+1,996	+2,239	+2,719
Worst 5% of runs	−640	−740	−1,460
Chance of finishing in loss	9.0%	11.2%	19.5%
Median drawdown	1,120	1,260	1,760
Chance of a 2,000 drawdown	7.8%	11.9%	34.8%

Reading the numbers honestly:

1. The doubled trades don’t win more often. Trades that followed 3 wins in a row won 49.9%, the same as any other. The streak says nothing about the next trade, because in this system each trade is independent
2. Doubling after 3 wins is almost the same as a fixed size about 12% larger. About 12% of trades get doubled, so the average size is 1.12R. We ran a fixed 1.12R for comparison: median +2,246 and a 12.3% chance of a 2,000 drawdown, very close. The extra profit is simply the result of risking more on average, not a new edge
3. Escalating makes the mean look good while the median doesn’t improve. The mean is high because a few runs hit long streaks and made a big lump. But the median stays around 2,100, while the chance of finishing in loss rises from 9.0% to 19.5% and the chance of a 2,000 drawdown rises from 7.8% to 34.8%, about 4.5 times

If the system has no edge at all (wins 50% at 1:1), every rule averages about zero. But the chance of a 2,000 drawdown is 29.7% for fixed size, 40.5% for doubling after 3 wins and 80.5% for escalating on a streak. More risk, nothing in return.

And this is the best case for sizing up, because it assumes trade quality stays the same. In real life, trades opened while overconfident are often outside the plan, so the edge on the big trades may be even lower. We didn’t simulate that part because there is no reliable number for it.

One exception, to be complete: if your system’s results genuinely cluster, for example a trend-following system whose winners tend to come in groups while the market trends, sizing up on a streak could help. But that has to be shown with several hundred trades, while watching for the counting bias above. For adjusting size by the equity curve, see equity curve trading.

Practical rules for after you win

1. Fixed risk per trade, regardless of the streak Set your risk per trade in the plan, for example 1% of the account, and use the same amount after wins and after losses. If you want size to grow with the account, recalculate on a fixed schedule, such as the start of each month, not after a big day. See growing position size with the account, and remember that higher risk per trade raises the chance of blowing up faster than you’d expect. See risk of ruin.

2. A cooling-off rule after a record day Write down in advance what a “record day” is, for example a day’s profit more than twice your average winning day, or a new account high. Then set the rule:

• No new trades for the rest of that day
• The next session uses the same risk % and the same maximum number of trades per day
• Before the next session, finish a written review of the winning trades

The “twice” figure is only a starting point. No research says which number is best. What matters is that the rule is written while you’re calm, not while you’re buzzing.

3. Review winners as hard as losers For every winning trade, answer four questions in your journal:

1. Was this trade in the plan, or opened on a feeling?
2. Was the size what the plan said?
3. Did the exit follow the rules, or did price happen to run in your favour?
4. While it was open, how close did price come to the Stop loss? See MAE and MFE

A trade that broke the plan and still made money is the most dangerous kind, because it teaches the brain that breaking the plan pays. Separate process from outcome with process over outcome and use the form from the trading journal template.

A trade that broke the plan but made money goes in the journal as a mistake, not a success.

4. Grow size on evidence, not on a feeling If you think your skill really has improved, set the conditions in advance: several hundred trades under the same rules, an average result per trade clearly above break-even, and drawdown still within the expected range. Then increase size in small steps.

Common mistakes

• Treating profit as “free money” and risking it more heavily than your deposit. Every dollar in the account is worth the same
• Sizing up after a win streak. If trade outcomes are independent, all you get is more risk, and escalating in several steps makes the bad side much deeper
• Trading more after a good day, taking trades outside your setup because “the market is going my way”. Every trade pays its cost
• Reviewing only the losers, so winners that came from luck are never checked and turn into habits
• Reading “after a streak” stats from a few hundred trades and concluding there’s a pattern, when the counting itself can skew the result
• Being unable to stop after winning, or feeling you must keep trading. If it happens often enough to affect your life, see warning signs of gambling and burnout and talk to a professional. In Thailand, the Department of Mental Health hotline is 1323