| Metric | Value | Copy |
|---|---|---|
| {{ row.label }} | {{ row.display }} |
A coin toss is a two-outcome probability trial. This tool shows the latest face, records every flip in the current session, and lets you switch between an ordinary browser draw and a repeatable seeded sequence.
That makes it useful for more than a one-time heads-or-tails answer. A quick choice can rely on the latest face alone, while a class demo or casual experiment can use the running counts, percentages, current streak, distribution chart, visible history, and JSON record to show what happened across many flips.
The advanced controls change either the probability model or the presentation around it. You can shift the chance toward heads or tails, run up to 5,000 fast flips, turn animation off, tune the visual spin, or enable optional vibration and beep cues when a heads result is revealed.
Those options are useful, but they also change what the session means. A short uneven run does not prove the model is biased, and a seeded run is only replayable when you start from the same seed state. Clearing the visible history with Reset does not rewind the seeded generator by itself.
Use the result as a simple decision helper, a repeatable teaching example, or a lightweight way to see long-run probability behavior. Do not use it to judge whether a physical coin is fair, and do not use system mode for security-sensitive randomness.
With Bias toward Heads (%) set to 0, the model is the familiar fair-coin case: heads and tails each have a theoretical probability of 0.5. The bias control moves that heads probability up or down before each draw, then clamps it into the valid range from 0 to 1. At +100, every recorded flip becomes heads. At -100, every recorded flip becomes tails.
RNG mode changes where the draw comes from. System uses the browser's built-in pseudo-random source. Seeded turns your seed text into a deterministic sequence, which is why the same seed can reproduce the same order when you begin again from the same seeded state. The page also flips once on load, so a fresh visit starts with one recorded result rather than a blank session.
Every resolved flip is appended to the session history. Flip Metrics, Flip Distribution, History, and JSON are all derived from that same stored sequence. Auto-flip xN runs without animation for speed and is capped at 5,000 flips. Animation length, spin turns, tilt, reduced-motion handling, vibration, and beep settings change the reveal experience but not the stored result. Vibration and sound only fire when Celebrate Heads is on, the outcome is heads, and the browser or device supports the feature.
Bias toward Heads (%), RNG mode, and an optional Seed set how each flip is generated.Flip Coin records one result. Auto-flip xN records a larger batch without animation.p(heads) = clamp(0.5 + bias / 200, 0, 1)
Heads if random draw < p(heads); otherwise Tails
Heads % = head count / total flips x 100
Current streak = number of identical latest results in a row at the end of the history
| View | What it shows | Export options |
|---|---|---|
Flip Metrics |
Total flips, heads, tails, percentages, current streak, RNG mode, and bias | Copy CSV, download CSV, DOCX |
Flip Distribution |
A donut chart showing how the session splits between heads and tails | PNG, WebP, JPEG, CSV |
History |
The flip order in reverse chronology so you can audit the run step by step | None |
JSON |
Inputs, derived statistics, and the full recorded history | Copy JSON, download JSON |
For a plain tiebreaker, keep Bias toward Heads (%) at 0 and leave RNG mode on System. Press Flip Coin and use the latest face as the answer. The earlier entries in History are still logged, but they should not outweigh the newest result when the goal is simply to decide.
Use Seeded mode when the run needs to be repeatable. That is the better choice for classroom demonstrations, test cases, and any situation where someone may want to reproduce the exact flip order later. If replay matters, keep the seed text with the result and restart from the same seeded state before comparing sessions.
If you want to show how probability settles over time, move from single flips to Auto-flip xN. A few flips can look dramatic by chance alone, even in a fair setup. A larger batch gives the percentages and donut chart more meaning, especially when you are trying to show the effect of a non-zero bias.
When the result needs a record, start with Flip Metrics for the summary, then use Flip Distribution for a visual check, History for order, and JSON for the full snapshot. The tool can export the metrics table, chart, and JSON state directly, which is useful when you need a quick handoff or a lightweight audit trail.
Reset. The tool clears the visible history and immediately records one new flip.Bias toward Heads (%) at 0 for a fair-coin assumption, or move it toward positive or negative values for a weighted demonstration.System for an ordinary browser draw or Seeded for a repeatable sequence, then enter a memorable Seed if needed.Flip Coin for one new result or enter a value in Auto-flip xN for a faster batch.Flip Metrics first. Then open Flip Distribution if you want a visual split, History if order matters, or JSON if you need the complete state.For a one-off choice, the latest face is the main result. For a run or demonstration, read the counts, percentages, and current streak together. The percentages describe what happened in this session so far. They do not predict what the next flip must be.
The current streak is especially easy to overread. It only counts how many times the latest face has appeared in a row at the end of the history. A streak can be interesting in a short run, but it is not evidence that the next result is due to swing back the other way.
| If you see this | Check this first | How to read it |
|---|---|---|
| A 70 or 80 percent split after only a few flips | Total flips and Bias toward Heads (%) |
Short runs can look uneven by chance. A small sample does not prove bias. |
| Only heads or only tails appear | Whether bias is at +100 or -100 |
At the extremes, the tool clamps the model so one outcome becomes certain by design. |
| The same seed does not replay the opening sequence | Whether you restarted the seeded generator or only pressed Reset |
Reset clears the visible log, but it does not rewind the existing deterministic sequence. |
| No vibration or sound on a flip | Celebrate Heads, the latest result, beep volume, and device support |
The cues are optional and only fire on a heads result when the feature is available. |
Switch RNG mode to Seeded, enter meeting-1 as the Seed, then press Reset. The new first recorded result is Heads. That gives you a documented tiebreaker you can replay later from the same seeded starting point.
Set Bias toward Heads (%) to 40, switch to Seeded, enter bias-demo as the seed, press Reset, then run Auto-flip xN with a value of 9. The session now contains 10 recorded flips in total, with 7 heads and 3 tails. It is a simple way to show how a 0.7 heads model can lean clearly toward heads without forcing every outcome.
Use Seeded mode with the seed class-demo and begin from Reset. The first five recorded results are Heads, Tails, Tails, Heads, Tails. If you press Reset and look at the next five recorded results without changing the seed or mode, the run continues as Heads, Heads, Tails, Heads, Tails. The fix is to restart the seeded state, not just clear the visible history.
The tool performs one flip on load so the summary, counts, and tabs are populated immediately. A fresh visit therefore starts with one recorded result instead of a blank session.
No. Reset clears the visible history and immediately records one new flip, so the session restarts from a one-result state.
A positive value raises the chance of heads above 0.5. A negative value lowers it below 0.5. The endpoints +100 and -100 clamp the model to all-heads or all-tails behavior.
Because the seeded generator keeps advancing as flips are recorded. Clearing the visible history does not reset that internal position. To replay from the start, begin from the same seeded state before the next run.
No. The tool is suitable for casual decisions, demonstrations, and lightweight experiments. It is not a security-oriented random service and it does not certify fairness for formal gambling or regulated use.