Skip to content
DiceSim

Learn · Simulator

How to read a dice simulation

A dice simulation gives you a balance chart and a panel of numbers. This guide explains how to read a dice simulation in DiceSim: what each statistic measures, how max drawdown, loss streaks and the efficiency score are defined, what a verified run is, and why one good-looking run tells you less than it seems. The definitions match what the simulator computes.

The stats panel

The panel next to the chart updates while the run is going. In order:

  • Rolls: the number of bets placed so far, with the total you asked for underneath. A run can stop early when the script calls stop() or the next bet is larger than the balance.
  • Win rate: wins divided by rolls, with the win and loss counts and the chance the script was betting at. Compare it with that chance. At 49.5% over 10,000 rolls the win rate has a standard deviation of about 0.5 percentage points, so a gap of that size is ordinary luck and says nothing about the strategy.
  • Net profit: balance plus vaulted funds minus the starting balance, with the same figure as a percentage of the start (ROI). If the script calls resetstats(), profit is re-based to 0 at that point.
  • Balance and Vaulted: the funds still in play, and the funds moved out of play with vault().
  • Wagered: the sum of every stake. It is the volume the house edge works on, so wagered × edge / 100 is the expected loss of the run.
  • Max drawdown, max win streak, max loss streak and current streak: covered below.
  • Elapsed and Speed: wall-clock time and rolls per second. They describe your machine, not the strategy.

A quick luck check

Compare the net profit with what the house edge predicts for the volume the run placed:

luck ≈ net profit + wagered × edge / 100
Positive = the run did better than average

A strategy with a profit of 0.002 after wagering 0.5 at a 1% edge was 0.007 better than the average outcome. That gap is the part of the result that came from the dice. The house edge guide explains why the expected part depends only on volume.

Max drawdown

The engine tracks equity, which is the balance plus everything vaulted. Max drawdown is the largest drop in equity from any earlier peak to a later low point, shown as a negative number and as a percentage of the starting balance. Because vaulted funds count toward equity, moving money to the vault is not a drawdown, and a loss of the active balance after vaulting is measured against the full peak.

Drawdown shows how bad the run got on the way, which the final profit does not. A run that ends +5% after being down 60% at one point needed most of its bankroll to get there. With slightly different rolls it would have gone bankrupt.

Loss streaks

Max loss streak is the longest run of consecutive losing bets. For progressions it is the most important number on the panel, because ruin comes from one streak longer than the bankroll can pay for. Compare it with the max survivable streak N from the calculator.

The longest losing streak in n rolls is typically close to log(n · p) / log(1 / q), where p is the win chance and q = 1 − p (Schilling, "The Longest Run of Heads", The College Mathematics Journal, 1990). At 49.5% over 10,000 rolls that is about 12.5 losses. The classic martingale with a base bet of 1/10,000 of the balance survives N = 13, which is why its risk of ruin over 10,000 rolls is about 49.7%. A run whose max loss streak stayed well below N never met the streak that ends the strategy.

The current streak shows positive numbers for wins and negative numbers for losses. The max win streak matters for positive progressions such as Paroli, which raise the bet after wins.

Efficiency

Efficiency appears on logged runs, the dashboard and the leaderboard. It is net profit divided by the size of the max drawdown:

efficiency = net profit / |max drawdown|
9999 when there was no drawdown and the run made a profit

A value of 2 means the run made twice its worst peak-to-trough drop. A value below 1 means it went through a drop larger than its final profit. Negative values are runs that lost money. It rewards smooth equity curves and penalizes runs that got lucky after a deep hole.

The Auto-Tuner ranks parameter sets with a related fitness score, efficiency multiplied by the natural log of the rolls survived, so that a run which lasts longer scores higher at the same efficiency.

Verified runs

Every simulation uses a server seed, a client seed and a nonce, and the rolls follow from them. When you submit a run to the leaderboard, the server replays the script with the same seeds and settings and checks that the numbers match. A run that matches is marked verified.

The server replays runs of up to 250,000 rolls. Longer runs are still saved to your account but cannot be verified or ranked. A verified run shows that the numbers are real for those seeds. It does not show that the strategy will do the same on other seeds.

Why one run proves little

One run is one sample from a wide distribution. The expected value and variance guide shows that a flat bettor at 49.5% is still ahead after 10,000 rolls in a noticeable share of cases, purely by chance. Any strategy will produce some winning runs.

The stress test

The Stress Test tab runs the same script and settings on 5 workers at once, each with its own random seed pair. The chart overlays all 5 balance lines with their median. The summary shows the best and worst run, the mean and median profit, the mean drawdown and how many runs went bankrupt. A table lists each run's seed, rolls, profit, max drawdown, longest loss streak and how it ended.

Read the worst run and the median first. If 2 of 5 runs went bankrupt, the strategy is close to a coin flip for that length, however good the best run looks.

The Bets tab

The Bets tab lists every bet of the main run (the most recent 1,000,000 on very long runs) with its nonce, roll, target, stake, payout, profit and balance. You can filter to wins or losses, jump to a bet number, jump straight to the longest losing streak and export the list as CSV. Looking at the bets around the biggest streak shows how large the stakes got and how close the balance came to the end.

What a good-looking run hides

A smooth, rising balance chart can hide several things. Check these before trusting it:

  • The run was shorter than the expected wait for a fatal streak, so the streak never had a chance to appear.
  • The max loss streak was below the survivable streak, which means the strategy was never tested at its weak point.
  • The max drawdown was close to the starting balance, so the run nearly ended before it recovered.
  • The wagered amount was large, so the expected loss was large too and the profit came from luck on top of it.
  • Other seeds give different results. One seed is one draw, and the stress test shows the others.

The risk of ruin guide shows how to estimate the chance of the streak that the run did not meet.

Try it

FAQ

What is max drawdown in a dice simulation?

It is the largest drop in equity (balance plus vaulted funds) from a previous peak, shown as a negative number. It measures the worst stretch the run went through, even if the run ended in profit.

What is the efficiency score?

Efficiency is net profit divided by the size of the maximum drawdown. A value of 2 means the run made twice its worst peak-to-trough drop. If there was no drawdown and the run made a profit, DiceSim shows 9999.

Why does my win rate differ from the win chance?

The win rate is wins divided by rolls, so it varies from run to run. It moves closer to the chance as the number of rolls grows. At a 49.5% chance the standard deviation of the win rate after 2,500 rolls is about 1 percentage point, so a gap of that size is ordinary.

What is a verified run?

A run the server replayed from its recorded seeds to check the numbers. Only runs of 250,000 rolls or fewer with recorded seeds can be verified, and a run that fails the replay is removed from the leaderboard.