Does Going First Matter in Commander?
Yes. In a 4-player pod the first seat wins 29.2% of games against a 25.0% fair share, while the last seat wins only 21.5% . Every seat down the turn order wins less, and the edge does not even out in longer games.
Seat 1 over an even 25.0% share in 243,938 four-player games.
The seat 4 win rate, relative to fair share. Every seat in between falls in order.
Finished multiplayer games (3+ players) tracked on Playgroup.
Does going first matter in Commander?
Yes, and by more than most tables assume. Across 491,808 tracked multiplayer Commander games, the first player wins well above their fair share and every seat down the turn order wins less. In a 4-player pod the first seat wins 29.2% of games versus the 25.0% fair share, and the last seat only 21.5% .
That is roughly a 17% higher win rate than fair for going first, and about 14% lower for going last. The edge is small in any single game but it is consistent, it holds across pod sizes, and, as the length data below shows, it does not wash out over a long game. The median game here runs 9 rounds and about 58 minutes .
Win rate by seat, for every pod size
Single-winner games only, so draws are excluded and the seat rates in each pod add up cleanly. Seat 1 went first; play then proceeds clockwise. The advantage is monotonic: first beats second beats third, all the way down.
243,938 single-winner games. The dashed line is an even 1-in-4 share.
167,052 single-winner games. The dashed line is an even 1-in-3 share.
34,512 single-winner games. The dashed line is an even 1-in-5 share.
Bars are win rate; the dashed line marks an even share. Error bars are 95% confidence intervals, tight because the samples run into the hundreds of thousands.
How we know this is not just noise
A win rate on its own cannot tell you whether a gap is meaningful or a small-sample wobble. To settle that we use the odds of winning. A seat that wins 20% of its games has odds of 20 to 80, or 0.25. We then compare each seat to the first seat with an odds ratio : the first seat is the baseline of 1.0, and a seat with an odds ratio of 0.66 has 34% lower odds of winning than going first. This is exactly what a logistic regression reports, and with hundreds of thousands of games the result is statistically overwhelming.
| Seat (4-player pod) | Win rate | Odds of winning vs first seat |
|---|---|---|
| Seat 1 (first) | 29.2% | 1.00 baseline |
| Seat 2 | 25.7% | 0.84 (0.83 to 0.85) |
| Seat 3 | 23.6% | 0.75 (0.74 to 0.76) |
| Seat 4 (last) | 21.5% | 0.66 (0.66 to 0.67) |
The value in parentheses is the 95% confidence interval, the range the true odds ratio almost certainly falls in. Every seat sits comfortably below 1.0 and nowhere near it, so the turn-order effect is real, not a sampling fluke. The same monotonic pattern holds in 3 and 5-player pods.
Does it even out in long games?
"Turn order stops mattering once the game goes long" is the common defense. It does not hold. Slicing 4-player single-winner games by length, the first-to-last gap is smallest in average-length games and biggest in the grindiest ones.
| Game length | Games | Seat 1 | Seat 4 | Gap |
|---|---|---|---|---|
| 6 rounds or less | 37,138 | 30.8% | 21.4% | 9.4pp |
| 7 to 9 rounds | 106,892 | 28.3% | 22.2% | 6.1pp |
| 10 to 13 rounds | 84,372 | 29.5% | 20.9% | 8.6pp |
| 14 or more rounds | 15,202 | 30.7% | 20.0% | 10.7pp |
In the grindiest games (14+ rounds) the first-to-last gap is 10.7 points, wider than in any shorter bucket. Acting first every turn cycle compounds, it does not decay.
Does a mulligan cancel it?
In 1v1 the first player skips their first draw to compensate. Does starting down a card fix multiplayer? A mulligan is the natural experiment: a first player who mulligans starts with fewer effective cards. The answer is no.
Win rate by how many times the player mulliganed, seat 1 versus seat 4.
| Hand kept | Seat 1 | Seat 4 |
|---|---|---|
| Kept 7 |
29.3%
n=209,697
|
21.6%
n=209,852
|
| One mulligan |
29.9%
n=22,408
|
22.1%
n=22,240
|
| Two or more mulligans |
27.0%
n=11,838
|
18.8%
n=11,841
|
A seat-1 player who mulliganed once wins as much as one who kept seven, and even two-plus mulligans deep still beats fair share and beats seat 4's kept-seven rate.
On Playgroup Live , our free in-browser virtual tabletop, the London mulligan is enforced by the software: the first mulligan is a free fresh seven, and from the second you physically bottom real cards.
| Hand | Seat 1 | Seat 3 |
|---|---|---|
| Kept 7 |
40.2%
±2.98 · n=1039
|
26.8%
±2.74 · n=1006
|
| One mulligan (free fresh 7) |
38.9%
±4.08 · n=548
|
27.9%
±3.65 · n=581
|
| Two or more (down a card) |
40.4%
±5.53 · n=302
|
23.6%
±4.8 · n=301
|
| Hand | Seat 1 | Seat 4 |
|---|---|---|
| Kept 7 |
28.6%
±3.63 · n=595
|
22.1%
±3.39 · n=575
|
| One mulligan (free fresh 7) |
29.9%
±4.83 · n=345
|
25.0%
±4.58 · n=344
|
| Two or more (down a card) |
25.5%
±6.04 · n=200
|
22.6%
±5.52 · n=221
|
Live samples are hundreds of games, not hundreds of thousands, so the confidence intervals are wide. Read them loosely. Even so, a first seat forced down a real card still wins at or above fair share.
Honest caveat: players mulligan their worst hands, so a mulligan row mixes "down a card" with "had a rough seven." That bias works against the first-seat advantage, yet the first seat still wins at or above fair share, which only strengthens the conclusion. The edge looks positional (you act first every cycle) rather than resource-based, so a one-card tax cannot touch it.
Who goes first in Commander (and which direction)?
Choose the starting player at random, with a die roll, a coin flip, or any method the table agrees on. There is no roll-off or bidding requirement in the rules. Play then proceeds clockwise, beginning with that starting player and continuing to their left.
The starting player skips their first draw step only in two-player games. In a multiplayer pod nobody skips a draw: everyone, including the first player, draws on their first turn. That extra card is part of why going first carries the edge shown above.
Method and caveats
- This is casual, self-reported Commander, not cEDH or tournament data. Real pods logging real games, mostly through the Playgroup life counter.
- Only games with exactly one winner are counted, so draws are excluded and seat win rates sum cleanly. Fair share is one divided by the pod size.
- Seating is not randomized by us. Players sit where they sit; we only record who went first.
- The seat and length numbers run into the hundreds of thousands of games, so confidence intervals are tiny. The Playgroup Live mulligan cells are much smaller; their intervals are shown and are wide.
- Computed 20 Jul 2026 from 491,808 finished multiplayer games.
Turn Order FAQ
Does going first matter in Commander?
How much is going first worth in Commander?
Does skipping the first draw fix the turn-order advantage?
Who goes first in Commander, and which direction does play go?
Do you draw a card on the first turn in Commander?
Is the turn-order advantage statistically significant?
Deal yourself the first seat
Playgroup Live is a free in-browser virtual tabletop for 2 to 6 player Commander. Randomized seating, enforced mulligans, and every game feeds the data on this page.