fair diceprovably fairbackgammon apps
Almost always, no. The rolls that feel rigged are usually honest randomness read through human memory: doubles arrive one roll in six, streaks are normal, and the losses stay with you while the wins fade. The harder truth sits underneath. With a closed app you cannot verify any of this, so the real question is not whether an app cheats. It is whether the app can prove that it does not.
Every backgammon app hears the same two complaints, and players have made them for as long as dice have rolled. The opponent gets too many doubles. The opponent always finds the number they need. The loser said it in the cafes of Istanbul when the dice were bone and the table was wood. It deserves a serious answer, not a shrug. Here is one.
Why does the app roll so many doubles?
Because doubles are common, and the ones that beat you are memorable. Two dice can land 36 ways. Six of those are doubles. One roll in six comes up paired, and that single fraction powers almost the entire complaint.
- A typical game runs 50 to 60 rolls across both players. Eight to ten doubles at the table is not a hot streak. It is the average game.
- Back-to-back doubles land one pair in 36. Play a few games in an evening and you should see it more than once.
- Three doubles in a row is one chain in 216. Play regularly and you will meet it. Arithmetic, not a glitch.
- Any single number your opponent needs shows on at least one die 11 times in 36, nearly one chance in three.
The bar tells the same story. Close two points of your home board and your opponent still enters about eight times in nine. Leave a blot within direct range and the number that hits it lands about one time in three before combination rolls add more. When the numbers they need are that available, "they always roll what they need" is simply what fair dice look like from the losing side of the board.
Why do the dice only feel unfair when you lose?
Because memory keeps a crooked ledger. Psychologists have measured this for decades: a loss weighs roughly twice as much as an equal win. So the 6-6 that wiped out your prime stays with you for a week, while your own 6-6 felt like justice and was forgotten by dinner. Nobody remembers the night the dice were average, and nobody posts about it either. That is why every rigged-dice thread reads the same, and almost none of them come with a tally.
This is not a flaw in you. Attention stores the roll that changed the game and discards the fifty that did not. Players were accusing dice of treachery when dice were carved from bone. The app inherited the suspicion the day it drew its first checker.
Could a backgammon app actually cheat?
Yes, and honesty requires saying so plainly. The tumbling cubes on your screen are an animation. The numbers come from code on a server you cannot see, and nothing physical constrains them. Most studios have no motive, since a rigged game is a dead reputation waiting to surface, and when players run counting tests on big apps the totals generally come back ordinary. But "probably not" is trust, not proof. With a closed app there is no experiment you can run that settles the question, because the architecture keeps the dice out of reach. So the honest division is not cheating apps versus honest ones. It is dice you can check versus dice you must take on faith.
You cannot audit a feeling. You can audit a receipt.
How can you test the dice yourself?
Before anyone reaches for cryptography, run the paper test. It settles most suspicions in a single evening.
Set the expectation first. Write down what fair looks like before you play: one roll in six is a double, for both sides. Deciding what counts as suspicious after you have lost is how bias wins.
Tally instead of remembering. For one full session, count every roll and every double, yours and your opponent's. Paper beats memory, because memory is the instrument that was misreporting in the first place.
Read the result like a statistician. In 120 rolls, expect about 20 doubles. Anywhere from the low teens to the high twenties is ordinary swing. Only a large gap that survives many sessions means anything at all.
Two things happen when players actually run this test. First, the opponent's doubles land near the base rate almost every time, and it is their own count that surprises them. Second, the test reveals its own ceiling. Totals can be clean while timing is dirty: a determined cheat could serve a perfectly fair share of doubles and still deliver the right one at the worst moment for you. Counting cannot catch that. Receipts can.
What does Orion publish for every game?
A receipt, in three parts. Before the first roll, our server commits to its dice: it publishes a SHA-256 fingerprint of a secret seed while the game is still unplayed. During the game, every roll is derived from that seed with HMAC-SHA256, one roll at a time, in order. After the last move, the seed itself is revealed.
That structure makes the dice checkable. Hash the revealed seed and it must reproduce the fingerprint published before anything was rolled, which proves the dice were fixed before either player moved. Re-derive the rolls and they must match the game you actually played, which proves nothing was swapped along the way. Change a single die anywhere and the chain refuses to close. The verifier at orion.arthea.ai/verify recomputes all of it in your browser using Web Crypto, with no account and nothing to install. And because the server validates every live move at source, an opponent cannot slip an illegal play past you while you are busy watching the dice.
We built it this way because "trust us" is not an answer we would accept from anyone else. The door at orion.arthea.ai/go carries the App Store and Google Play links and always shows what is available right now. The verifier went live first, on purpose. The proof should be waiting before the players arrive.