Euchre by the Numbers: When to Order Up, Pass, or Go Alone

Most euchre strategy is folk wisdom: “three trump and a bower,” “always call next,” “never go alone without both bowers.” We wanted numbers. So we built a euchre simulator, played 1,000,000 deals, and at every bidding decision played the hand out three ways, once calling, once passing, and once going alone, to see which choice actually scores more.

This page shares what we found, how we did it, and where the method falls short. If you want the answer for one specific hand, use the hand evaluator.

Key Findings

1. The deal is worth about half a point

Across all hands, the dealer’s team outscored the other team by 0.45 points per hand. The dealer can pick up the turned card and throw away their worst card, and their partner can order it up knowing it lands on their own side. Over a game that’s a real edge. It’s why the deal rotates, and why score management near the end of a game should account for who deals next.

Who ended up calling trump:

Trump was set in round one on 74.2% of deals, in round two on 25.2%, and 0.6% were thrown in.

2. How much you need to order up depends on your seat

The same hand is a call from one seat and a pass from another. The minimum number of trump (counting the picked-up card for the dealer) at which ordering up beat passing on average:

“Not reached” means that, in this seat, no number of trump in that column beat passing on average.

The pattern matches the logic of the game. First and third seat hand the turned card to an opponent, so they need their own tricks. The dealer’s partner gives the card to their own team, so ordering up is much cheaper. That’s the strongest “call lighter than you think” result in the data. See the full tables below, and the bidding strategy guide for how to apply this at the table.

3. Call next against the dealer, cross as the dealer’s partner

In round two, when the turned card has been turned down, the suit of the same color is called next. The other two suits are cross. Holding the same number of trump, here is how often each call was made:

The old advice holds, and the data shows why it depends on the seat. First and third seat (the dealer’s opponents) do much better calling next. The dealer passed up a card of that color, which suggests the dealer’s team is weak there. The dealer’s partner does better calling a cross suit (“reverse next”), since their own side turned down that color. For the dealer, next was better with two trump and cross was better with three. More on this in calling next.

4. Going alone pays more often than players think

A failed loner that still takes three or four tricks scores 1 point, the same as a normal make. So the only real cost of going alone is the march you might have made with your partner’s help, plus a slightly higher euchre risk. These are the hands (as maker, after any pickup) where going alone most outscored playing with a partner:

Read these loner numbers as an upper bound. Our computer defenders are weaker against a lone player than an experienced human pair would be. The direction is clear, though: with both bowers, a third trump, and an off-suit ace, going alone is usually right. Without the ace, it’s a coin flip, so play with your partner. See when to go alone.

Full Tables: Round One, Every Seat

Each row is a type of hand, with results averaged over every simulated hand of that type. Call and Pass are the average net points for your team (your points minus theirs) after that choice. Rows are sorted from the strongest calls to the clearest passes. The turned-up card is excluded from the hand except for the dealer, whose hand is shown after picking it up.

How We Did It

1. The engine. A complete euchre rules engine: 24-card deck, two bidding rounds, the bowers, follow-suit rules, and standard scoring (march 2, loner march 4, euchre 2). There’s no Stick the Dealer: if everyone passes twice, the hand is thrown in.
2. The players. All four seats are played by the same computer players that power our free euchre vs computer game. They estimate tricks from trump, bowers, aces, and voids; pull trump when they’re the makers; don’t overtake a partner who’s already winning; and track which cards have been played.
3. Branching. Every time a player faces a bidding decision (because everyone before them passed), we copy the deal and play it out three times: once where they call, once where they pass and the auction continues, and once where they go alone. Comparing the branches removes the luck of the deal, because the only difference is the decision.
4. Calibrating the opponents. A naive computer player bids too much in some spots, which makes passing look artificially good: you “trap” an opponent who then gets euchred. To avoid that, we used iterated best response. After each 200,000-deal run, the computer players adopted any call-or-pass decision the data clearly favored (by more than about two standard errors), and we re-ran. After three iterations the results stopped changing. Then we ran the final 1,000,000-deal study.
5. Grouping. Hands are grouped by seat, number of trump, which bowers are held, and number of off-suit aces. Groups with fewer than 1,000 hands are left out of the tables.

Limitations

• Computer card play isn’t perfect. It’s solid, not expert, so treat absolute make rates as approximate. Comparisons between choices on the same deals (call vs pass) are much more reliable.
• Hands are grouped coarsely. A guarded King and a bare King count the same, so your specific hand may be a little better or worse than its group.
• The score is ignored. Every hand is valued on its own. At 9–9, or when the opponents are one point from winning, the right choice can change (score management).
• Loner defense. As noted above, the loner results overstate the value of going alone somewhat.
• Rules. House rules like Stick the Dealer, Canadian Loner, or defend-alone would change some numbers.

Download the Data

The data is free to use with a link back to this page:

• Summary tables (findings.json)
• Whole-game statistics (game.json)
• Loner vs partner (loner.json)
• Round 1 decisions by hand type (round1.json)
• Round 2 decisions by hand type (round2.json)

If you’re a researcher, writer, or developer and want the simulator or a custom cut of the data, get in touch.