Seats 3 and 4 Miss Catan's Top-Three Corner 43% of the Time. The Draft Is Not the Reason

By the Catan Trainer team · November 2026 · 8 min read · raw data (JSON)

Everyone who draws seat 4 says the same thing: the good corners were gone. We can check that, because when Catan Trainer grades a first settlement it ranks it only against the corners that were still legal at the moment it was placed. Every corner the earlier seats took is already out of the comparison. Do it that way and seats 1 and 2 land in the top three on about 73% of openings, while seats 3 and 4 land there on about 57%. The corners were not gone. The later seats are choosing worse from what was left, and a first settlement outside the top three is worth 35.1% win rate instead of 45.3%.

How we measured it

The source is 457 base four-player games imported from colonist.io by Catan Trainer users through our Chrome extension, with our own games excluded. Of those, 412 have a final result, 153 have a full AI review stored, and 456 produced a usable seat and first-settlement read. For every one of them the trainer rebuilds the board, replays the snake draft up to the moment the player placed, enumerates every corner that was still legal at that instant, and scores all of them with the same evaluator the practice trainer uses. Then it asks where the corner the player actually took landed in that list. One caveat matters: this group wins 41.5% of their four-player games against a 25% baseline, so these are stronger-than-average players. On Colonist at large the gap below is probably wider, not narrower.

That method is the whole point. "Top three" never means the top three on the opening board. It means the top three of what that seat could still legally take. Seat 4 is not being graded against the 13-pip corner seat 1 opened on.

The gap

Pool the seats and it is 73.1% for the first half of the draft against 56.7% for the second, a gap of 16.4 points. (We get those two by recombining the per-seat rates and counts above, so they carry a little rounding.) Notice there is no gentle slope. Seats 1 and 2 are within a point of each other, seats 3 and 4 are within three points of each other, and the cliff sits in the middle of the draft. Whatever changes, changes between the second pick and the third.

What missing the top three costs

Rank is only interesting if it predicts results, so here are the same games split by where the first settlement landed, against final placings.

Group the top two rows and you get 45.3% over 258 games. Group the bottom two and you get 35.1% over 154. Being inside the top three is worth a little over ten points of win rate, and being exactly first is worth nothing extra: the best corner and the second-or-third corner are a coin flip apart, and if anything the second-best corner is ahead. Perfectionism is not the lesson here. Showing up in the top three is.

So what does the seat gap actually cost? Multiply it out: 16.4 points of extra top-three misses, each worth about 10.2 points of win rate, works out to roughly 1.7 points of win rate for seats 3 and 4. That is a floor, not a measurement. It only counts the in-or-out-of-top-three split and ignores how far outside the misses land, and we do not publish win rate by seat, so we are not going to pretend we measured it directly. Call it one to two points, free, for a habit change that costs you nothing but attention.

Why it happens at the midpoint of the draft

Three things change when the draft turns over, and only one of them is about corners running out.

The question changes shape. Seat 1 answers an easy question: of all 54 corners, which is best? That answer is usually obvious, which is exactly why 73.6% of seat 1 picks are top three. Seat 3 answers a harder one: of the corners left, which fits what the two players after me will leave me? Seat 4 answers the hardest version, because seat 4 picks twice in a row with nobody in between. Our read is that seats 3 and 4 are facing a pair question and answering it as two separate corner questions, one at a time, the way seat 1 gets to.

Supply collapses faster than people expect. On 1,912 graded practice boards the best available first corner is a median 12 pips, and the average board offers 7.8 corners at 10 pips or better and 3.7 at 11 or better. By the second round of picks that is down to a median best of 10 pips, with 1.6 corners at 10-plus and 0.6 at 11-plus. That measures the drop between a board's first round and its second, not between seat 1 and seat 4, so do not read it as the seat gap. Read it as the direction. The shelf empties, and it empties quickly enough that a seat 3 player running seat 1's mental routine will find the routine has stopped working.

Close calls get closer. This one cuts against us, so here it is up front. When five corners are within a pip of each other, the difference between third best and fourth best is almost nothing, and a player of identical skill will drop out of the top three more often purely by chance. Part of the 16-point gap is probably that. But the win-rate table does not care why a corner ranked where it did, and the players whose first settlement fell outside the top three won 35.1% of their games. The cost is real whether the cause is skill or variance, and the fix is the same either way.

What we can rule out is the excuse. The ranking is computed against what was available, so "the good spots were taken" is already priced in. See the full placement data study for the rest of what the same 457 games say.

The fix: rank pairs, not corners

If you are seat 3 or 4, stop asking which corner is best. Ask which two corners you can still own, and pick the first one so that the second one survives.

Seat 4 has the cleanest version of this, and most seat 4 players throw it away. Your two settlements are consecutive. Nobody picks between them. You are the only seat that gets to choose a pair with certainty, which is worth more than the single best corner on the board, and the way to waste it is to take the highest-pip corner left and then take the next-highest-pip corner left.

Here is what that looks like in words. Brick is the board's shortage, two brick hexes on a 5 and a 9. Three settlements are down, all on the ore-and-wheat side. Two corners are open to you. Corner A is ore 8, wheat 6, sheep 11, which is twelve pips of exactly the resources everyone else already has. Corner B is wheat 5, brick 9, wood 10, which is eleven pips and the only brick left in reach. The seat 4 trap is to take A because twelve beats eleven, then take the best remaining corner, which on that side of the board is more ore and more wheat. You finish with twenty-three pips, four of the five resources, zero brick, and a road network you cannot build.

Take the pair instead. A plus B is twenty-three pips across all five resources with the board's scarce brick in your hand, and the one pip you gave up on the first pick comes straight back the first time somebody needs brick for a road. If pips are not yet automatic for you, the dice probability and pip guide has the full table, and the starting placement walkthrough runs the whole sequence in the order you should run it under a timer. The same reasoning is why ignoring the scarce resource is the most expensive habit we measure.

If you are seat 1 or 2, your trap is the other one

Seats 1 and 2 nail the first corner and then drift, because their second settlement arrives after six other settlements and whatever they planned is usually gone. Across all 456 openings the second settlement is a median 9 pips against 11 for the first, and 18.4% of second picks left two or more pips on the table against 13.4% of first picks. Early seats should place their first settlement so that the second has two or three acceptable landing zones, not one.

The twenty-second version

1. Rank the corners that are legal now. The ones that are gone are not your problem and are not in your grade either.
2. If you are seat 3 or 4, name your second corner before you place your first. If your first pick only works when one specific corner survives, it is the wrong first pick.
3. If you are seat 4, take the pair. Two settlements chosen together beat one perfect corner and one leftover.
4. Top three is the target, not first. First is worth nothing over second; outside the top three costs ten points of win rate.
5. Count what the board is short of before you count pips. That is where seats 3 and 4 lose the pair, every time.

None of this makes seat 4 better than seat 1. It makes seat 4 as good as seat 4 is supposed to be, which is most of what ranked Catan is. Our Colonist.io strategy guide covers what to do with the rest of the game once the opening is sound.

Keep reading: Catan placement mistakes, by the numbers · The scarce resource mistake · Starting placement, step by step · How to rank up on Colonist.io