Risk of Ruin Calculator: What Is the Chance of Reaching Your Loss Limit?

Here, “ruin” means reaching the loss limit you choose, not necessarily losing the whole account. With a 20% loss limit, the boundary stays at 80% of the starting balance. A close exactly on that boundary counts, as does a close below it.

Choose a win probability, a net win-to-loss ratio, risk per trade, a loss limit and a trade horizon. The calculator compares fixed cash risk with a percentage of current balance. The starting balance is normalized to 100%, so the probability does not change merely because you express the same scenario as a $1,000 or $10,000 account.

The model checks balance after each completed trade. It does not observe unrealized losses within a trade, margin liquidation or the changing rules of a trading programme.

The illustrative starting scenario uses a 45% win probability, a 1.5-to-1 net payoff, 2% initial risk, a 20% loss limit and 100 trades. Its first-touch probabilities are about 13.87% with fixed cash risk and 12.14% with current-balance risk. These are conditional model results, not measured failure rates or a suggested setup.

Each modelled trade has exactly two possible outcomes: a net gain of b risk units or a net loss of one risk unit. The win probability p stays the same on every trade. This deliberately simple description makes the calculation auditable; a real record with varied wins and losses contains information that these two inputs cannot preserve.

Include known costs in the net gain and net loss assumptions. Do not subtract the same commission or financing charge again after using net outcomes. If costs or execution losses vary, replacing them with a fixed amount also simplifies the distribution; the result does not capture the omitted variation.

There is no breakeven outcome in this model. Do not enter wins divided by all trades and silently treat every breakeven trade as a full loss. Likewise, counting partial closes as separate independent trades changes the unit being modelled. Define your observation consistently before using a journal as a starting point for a scenario.

The form accepts 1–1,000 trades, a win probability from 0% to 100%, a payoff from 0.10 to 10, risk from 0.01% to 10%, and an initial-balance loss limit from 1% to 90%. Decimal fields allow up to two places. These are tool limits, not trading recommendations. The fixed cash amount remains coverable in every still-active model state because the minimum allowed floor is at least as large as that amount.

The comparison changes one rule while keeping p, b, the starting balance, the limit and the horizon the same. Both approaches start with the same monetary risk. After that, they respond differently to profits and losses.

Fixed cash keeps the original loss amount unchanged. Its percentage of the remaining balance therefore grows after losses. Current-balance sizing reduces the cash amount after a loss and increases it after a gain. Neither rule moves the fixed floor.

A lower result in one comparison is not a universal ranking of the two methods. The result depends on the selected event, payoff, probability, risk and horizon. The calculator neither chooses a risk level nor maximizes expected growth. For converting an independently chosen monetary risk budget into a position quantity, use the position size calculator with the instrument’s specifications.

Imagine a starting balance of 100 and a floor of 60. With a fixed loss amount of 20 and equal gains, the sequence loss, loss, win, win would read 100 → 80 → 60 → 80 → 100 if it were allowed to continue. Looking only at the final 100 would miss the contact with the floor on trade two.

The first-touch calculation records the event on that second trade and removes that path from further active calculation. Later branches cannot restore it to the “no touch” group. This is why a probability of ending at or below the limit cannot replace the probability of ever reaching it during the horizon.

Any segment shown after first contact is a hypothetical continuation used to explain the counting error. It does not imply that trading could continue after an actual account restriction.

“No touch” is the complementary event: every observed close remained strictly above the chosen floor. Such a path can still end with a loss. No touch does not mean profit, absence of a large retreat from a later peak, or absence of an intratrade breach.

This separate teaching example starts at 100, fixes the floor at 60, uses four trades and gives a win and a loss equal probability. Each win pays one risk unit. The fixed cash amount is 20; the current-balance fraction is 20%. The unusually large risk makes the arithmetic visible and is not a recommendation or a preset for the main form.

There are 16 possible win/loss sequences, each with probability 1/16 because p is 50% and the outcomes are independent. Select a path in the explorer to see its first contact and its hypothetical continuation. W means win and L means loss.

Fixed cash reaches the floor in six paths: 6/16 = 37.5%. Current-balance sizing reaches it in two: 2/16 = 12.5%. Checking only the hypothetical ending values would count five and one respectively, or 31.25% and 6.25%. Those smaller numbers answer a different question.

The fractional path LLLW illustrates the same issue without landing exactly on the boundary: 100 → 80 → 64 → 51.2 → 61.44. Its first contact is on trade three. Finishing hypothetically above 60 cannot erase that contact. Outside this equal-probability teaching example, paths need not carry equal probability, so simply counting them would be wrong.

The input probability is a scenario assumption. A run of profitable trades does not establish a stable future p, and a sample average payoff does not describe the size or frequency of unusually large losses. Two strategies with the same win rate and average win/loss ratio can have different real loss-limit risks because their full outcome distributions differ.

• Independent outcomes: a loss does not change the chance of the next outcome. Clustering caused by shared exposures or market conditions is not modelled.
• Stable inputs: p, b and the sizing rule remain fixed for the whole horizon. No strategy switching, recovery staking or martingale is included.
• Two net outcomes: every win has the same relative size and every loss is a full modelled loss. Breakeven trades, outliers and variable partial-exit results are not separate states.
• Sequential closed trades: positions do not overlap. Deposits, withdrawals, interest on idle cash and external account adjustments are excluded.
• A fixed starting-balance floor: no trailing high-water mark, daily reset, profit target or time-based rule changes the stopping event.

A useful way to explore uncertainty is to change an assumption and compare the result: a lower win probability, a smaller net payoff or a longer horizon. Keep the other inputs fixed so you can see what changed. These comparisons are scenarios, not confidence intervals or estimates of how likely the assumptions themselves are.

The model’s arithmetic expectancy is p × b − (1 − p) risk units per trade. For p = 0.45 and b = 1.5, this is +0.125. A positive value does not remove sequences of losses or imply that the loss limit cannot be reached. It also does not by itself establish a positive compounded growth rate.

Maximum drawdown concerns the largest decline from an earlier peak to a later trough. That peak-to-trough definition is also used in the primary research on maximum drawdown. The present calculator instead watches a level fixed from the starting balance.

For example, start at 100, rise to 140 and then fall to 110. The account is still above a fixed floor of 80. Yet the decline from the 140 peak is about 21.43%. The two observations can both be true; changing a label from “loss limit” to “drawdown” does not make the events equivalent.

A programme may monitor open-position equity, adjust a trailing threshold, use daily limits or apply its own equality rule. This calculator has none of those mechanisms. Read the trailing drawdown guide to separate a threshold’s reference, its update schedule and the moment it is checked.

The model also does not reproduce the price path inside a trade. A trade that eventually closes at a gain could have crossed an equity limit before closing. Execution costs and slippage can make a realized loss differ from a planned stop amount. An after-close binary calculation cannot measure those omitted events.

The calculation starts with probability 1 above the floor. For each trade, it divides the probability still active at each state into a win branch with weight p and a loss branch with weight 1 − p. Any branch reaching or crossing the floor is added to the first-touch total and stops contributing to future active states.

This is an absorbing-boundary calculation using conditional probabilities. Cambridge’s Markov-chain notes explain hitting probabilities and absorbing states. The finite trading model here is our stated application of that framework; the sources do not establish the inputs for any real trading strategy.

Let B₀ be the initial balance, f the risk percentage written as a fraction, b the payoff, n the number of trades and w the number of wins. Fixed cash uses s = f × B₀. At a given state its balance is B₀ + s × [b × w − (n − w)]. Current-balance sizing gives B₀ × (1 + f × b)^w × (1 − f)^(n − w).

Although balance at a state depends on n and w, not every sequence arriving at that state has avoided an earlier contact. The calculation carries only the probability of paths that have stayed above the floor. Treating every end-state binomial probability as active would lose this essential history.

The results use the finite set of model states rather than randomly sampled Monte Carlo paths. Repeating the same inputs therefore gives the same result. Probabilities are still calculated with finite numerical precision and displayed with rounding; this is not an exact probability of real-world bankruptcy.

A familiar infinite-horizon gambler’s-ruin formula has narrower assumptions. For independent equal-sized gains and losses, starting k whole loss units above an absorbing lower boundary with no upper target, eventual contact has probability 1 when p ≤ 0.5 and [(1 − p)/p]^k when p > 0.5. Columbia’s derivation explains this case. It is not a substitute for a finite horizon, unequal payoff or a fraction that resizes after each trade.

Very small positive values are displayed as less than 0.01% instead of being rounded to a misleading zero. Values near, but not equal to, certainty are shown as greater than 99.99%. A genuine 0% can occur when contact is impossible within the selected model and horizon, for example when even all losses cannot reach the floor. That statement does not cover events or losses the model leaves out.

What does a 10% risk-of-ruin result mean?

Under the exact assumptions entered, the probability of at least one observed close at or below the fixed floor within the selected number of trades is 10%. It does not mean losing 10% of the balance, a 10% chance on every individual trade or a forecast based on your live account.

Can I calculate risk of ruin from win rate alone?

No. The payoff, size of the modelled loss, loss boundary, number of trades and sizing rule also matter. Even all of those inputs describe only this two-outcome scenario, not the full distribution of a real strategy.

Why do different calculators give different answers?

Check the event and the assumptions before comparing the percentages. A zero-balance target, a positive loss floor, a running-peak drawdown, an infinite horizon, a finite horizon and a terminal-balance event are different questions. Net versus gross payoffs and fixed versus resized stakes also change the model.

Does risking a percentage of balance make ruin impossible?

With a fraction strictly below 100% and exactly the bounded outcomes assumed here, a finite sequence leaves a positive balance. It can nevertheless reach a positive loss floor. Real gaps, costs, execution constraints and account rules are outside that simple multiplication model.

What if I do not know my future win probability?

Explore clearly labelled scenarios rather than treating a guess as an estimate. If historical records are used, keep the trade definition and net accounting consistent and acknowledge sampling uncertainty. This calculator does not fit, validate or predict the inputs for you.

Can this tell me the safest risk per trade?

It compares selected scenarios; it does not choose a suitable risk budget or certify safety. A small modelled first-touch probability can coexist with an inaccurate payoff assumption, missing tail losses, correlated positions or a different real account limit.