Hi-Lo Is in the Verifier Now, and Why It Was the Annoying One

Published 2026-06-24 · Tags: verifier, product

Hi-Lo verification needs the whole card chain plus your guesses, not one number. Full chain view, a higher/lower analyzer, and the ace rule explained.

Hi-Lo has been in the verifier in a basic form for a while: paste seeds, see the first card. That was never good enough and I knew it. This week it got the full treatment.

What was wrong with the old one

Most Stake Originals are one bet, one result. Dice is a roll. Limbo is a multiplier. You derive a number from the HMAC and you're done.

Hi-Lo isn't that. A single Hi-Lo game is a chain of cards, and each guess you make (higher, lower, skip) advances to the next card in the sequence. To verify a round you don't need one card, you need the whole chain, and you need to know which position in the chain each of your guesses landed on. The old verifier showed you card one. If your dispute was about card seven, tough.

What it does now

Enter the seed pair and nonce and you get the complete card chain for that round, laid out in order, with the position index on each card. Click any guess type on any card and it tells you whether that guess would have won and at what multiplier, using Stake's actual payout rules (ace is low, suits don't matter, "same card" is a push on higher/lower except at the ends of the range where it's a loss).

There's also a higher/lower analyzer, which is the part that took the time. Give it a seed pair and a strategy (always guess the statistically favored direction, say, or "higher below 8, lower above 8") and it replays that strategy across thousands of nonces and reports the realized win rate and the longest losing run. The point isn't to find a winning strategy; there isn't one. The point is to show, with your own seeds, that the best possible Hi-Lo strategy still lands under 100%. People find that more convincing than being told.

The ace thing

Quick note because it comes up every time: Stake's Hi-Lo ranks ace as the lowest card and king as the highest, and suits don't matter. The multipliers follow from that. Guessing "higher" on an ace pays almost nothing because almost every card qualifies; guessing "higher" on a king is the long shot. The verifier implements Stake's exact rules, including how a repeated card is scored, so if your mental model disagrees with the verifier, check the rule tooltip before you check the seeds. Nine times out of ten the disagreement is about the ace.

Why it took a while

The chain. Every other game's verification is a pure function of (seeds, nonce). Hi-Lo's is a pure function of (seeds, nonce, your sequence of actions), and modelling the actions correctly, including skips, meant building a small game engine rather than a formula. It also meant the analyzer couldn't reuse the deep-dive scanner we use for Dice and Limbo; it needed its own.

It's done. It's at /verifier/hilo. Blackjack, which has the same shape of problem (a chain of cards plus player actions), is next on the list for the same upgrade.