The talent picks and gold choices of one FlagDecker match

One match, start to finish, and the talent and gold calls you face along the way. The bots we watched made most of these choices badly, which turned out to be the most useful part.

Take your first talent before round 1 with the whole match in mind

You open the match with three talents on offer, usually from different categories, and no board at all. Every one of the 24 talents can show up here. That matters more than it seems, because the second pick before round 7 never offers the early ones: War Chest, Blood Tax, Street Market, Free Dispatch, Levy, Small Nations and Blitz. If you want any of those, now is the only time.

A few picks also reach into your shop. Levy removes Militia, Lone Stars removes Outposts and Treasury removes Coffers. Lone Stars will never be offered alongside Silk Road, Old Continent, Coastline, Long March or Empire. If the rules of the shop are new to you, the how to play page covers them, and the master guide covers how boards score.

Our habit is to treat this first pick as a gold decision. You cannot judge a scoring talent against a board you do not have yet, but you can count gold. With no economy talent you get 6 to start, then 6, 7, 7 and 8 a round from round 5. Spend it whenever you can and that is 114 gold and 38 card buys over 15 rounds.

Blood Tax is the one we reach for. It costs 8 HP on the spot and pays +4 gold every round after, which comes to 56 extra gold: 170 in total and 56 buys. The HP sounds expensive until you remember that losses in rounds 1 to 4 are capped at 3 HP each. Eight HP is roughly three of those early defeats.

One bot in our test matches made the case better than we can. It took Blood Tax, started round 1 on 24 HP, kept losing, and was down to 8 HP by round 7. By round 10 it had 2 HP and was sitting on 17 to 26 gold. Its second talent was Last Stand. It finished first, still on 2 HP, with Tajikistan, Kyrgyzstan, Uzbekistan, Afghanistan and China plus Grand Alliance and Crossroads. The board scored 126, and Last Stand's +12 at low HP made it 138.

We still only rate Blood Tax A, and only for players who plan their gold. Bots placed 3.26 on average with it, because they bank the money and forget to turn it into cards.

War Chest is the opposite deal, and we like it a lot less. You start with 10 gold and get 1 less every round after. That buys a third card in round 1. By round 5 you are level at 34 gold, from round 6 you are behind, and by round 15 you have 104 gold and 34 buys, 10 gold short of doing nothing. Bots placed 3.61 with it as a first pick, among the worst. The War Chest bot that did win bought its 3 cards in round 1, then took Sixth Seat at round 7 and finished at 22 HP with India, China, Myanmar, Laos, Bangladesh and Bhutan, Vanguard and Grand Alliance, for 127. The second pick won that game.

Interest is the patient option. At the start of each round you get +1 gold per 2 gold held, up to +5 once you hold 10. Keep a 10-gold bank from round 2 and you only buy 2 cards in rounds 2 and 3 instead of 4. It feels awful. By round 4 you are level on cards, from round 5 you are ahead (12 against 11), and you finish with 55 buys and 177 gold. A smaller 4-gold bank still gets you 47 buys.

Our favourite Interest bot never lost a single HP. It went from 32 to 32, sat on 20 to 30 gold from round 8, and won with Brazil, Argentina, Paraguay, Bolivia and Uruguay plus Grand Alliance and Crossroads, 120.

Spend rounds 1 to 4 cheaply, because losses only cost 3 HP

The early fights feel urgent. They are not. The loser takes a base amount plus 1 for every 4 points of score difference, and in rounds 1 to 4 the base is 1 and the cap is 3. Cards cost 3, sell for 1 and a refresh is 1 gold. The shop has four slots.

Street Market makes cards cost 2 in this window. A normal player buys about 8 cards in rounds 1 to 4, so it is worth about 8 gold, all of it up front. Free Dispatch makes your first 2 refreshes free every round. That is worth up to 30 gold, but only if you would have refreshed twice every round anyway. We think it is better than it looks for anyone hunting one particular border.

Blitz doubles your total in rounds 1 to 4. A round-4 board of Germany, France, Switzerland and Italy scores 24, or 48 with Blitz, which looks great on the scoreboard. With losses capped at 3 HP, though, the most it can swing is 3 HP a round. We put it in C.

Levy gives +2 for every base-1 card, and the base-1 cards are the big famous countries such as China, Russia, USA and Brazil. On China, Pakistan, Iran, Mongolia and India it adds 10. Remember it bans Militia.

Small Nations doubles your smallest country by area. On the boards we tested that was usually +4 or +5, and +1 on that all-base-1 Asian board. Fine, never exciting.

Hold ten gold through the middle rounds if you took Interest

From round 5 income settles at 8 a round and the shop grows to five slots, then six from round 9. Losses now cost a base of 2, capped at 6. Tier 2 effect cards have been around since round 3, and tier 3 joins in round 6.

This is where Interest players crack. The shop shows a good card, you have 11 gold, and buying it drops you under the 10 that pays the full +5. Hold. If you skipped Interest at round 1 and take it as your second pick, it only overtakes plain income on cards at round 10 (26 against 24) and ends at 153 gold and 47 buys. Still decent, much less exciting.

Treasury lives off the same pile. From round 7 it adds +1 to your score per 2 gold held, up to +12 at 24 gold, and it bans Coffers. Bots placed 3.50 with it as a first pick and 3.30 as a second, which is about what you would expect from a talent that rewards hoarding when they were not trying to hoard.

Grand Bazaar is about access. Effect tiers open 2 rounds early, tier 2 in round 1, tier 3 in round 4 and tier 4 in round 7, and the shop is 5 wide from round 3 and 6 wide from round 7. It gives you more to look at and no extra gold to pay for it.

Royal Court guarantees a base 5+ card in the shop from round 7 and gives your base 5+ cards +2. That is +10 on a board of five tier-5 flags, +8 on the Central Africa chain and nothing on a tier-1 board. Take it when your board is already rare flags.

Use the round 7 pick to fix what the shop gave you

By round 7 you have a board, so the second pick is a scoring pick. It is also your chance to repair the first one.

We watched one bot take Silk Road first and then never see an Asian link. It built Belarus, Poland, Lithuania, Russia and Latvia instead. Russia counts as Europe, so Silk Road added nothing at all. At round 7 it took Sixth Seat, filled the sixth slot in round 9, and finished on Belarus, Poland, Lithuania, Ukraine, Russia and Latvia with Grand Alliance and Crossroads for 160. Without Ukraine the same board scores 129. It won the match.

That is why the two slot talents sit at the top of our list. Sixth Seat gives you six card slots from round 9. The Central Africa chain plus Chad scores 88 instead of 74, and the chain cards grow with it: Vanguard is worth +22 and Imperium +24 on six. Privy Council gives you a third effect slot from round 8. Central Africa with Grand Alliance and Sovereign is 134, and Imperium in the third slot makes it 154. In the bot data they were the two best second picks, Sixth Seat at 2.89 average place with 30% wins and Privy Council at 2.93 with 28%.

The link talents are the big swing if the shop cooperates. Silk Road pays +5 per Asian link, which is +25 on China, Pakistan, Iran, Mongolia and India, taking 36 to 61. Old Continent pays +4 per European link, +36 on Russia, Belarus, Poland, Lithuania and Latvia. As first picks they had the best bot numbers we saw, 2.67 and 2.76 average place and 41% wins each. Just do not take one hoping the links arrive later. The Silk Road bot above is what that looks like.

Coastline is the forgiving version: +1 per link on each coastal country, so +14 on that Baltic board and +12 on Central Africa. Empire doubles your largest country and its links, +12 on Central Africa, but the biggest countries are tier 1, so it often ends up doubling a 1. Regalia doubles your highest base card, usually about +5.

Long March is the one we would leave alone. It multiplies the chain bonus by 1.5, but chains under 4 get nothing, so it costs you 5 on a 3-card chain and adds 9 on a five-card one. It had the worst first-pick number in the bot data, 3.73. Exile Court is not much better. People without their homeland get half the homeland bonus, rounded up, so Qin Shi Huang without China goes from 37 to 40. Bots placed 3.64 with it first and 3.60 second.

Here is where everything lands for us:

The bot numbers come from 4,000 matches where talents were picked at random, and the bots play one fixed style. A player who plans will get more out of Interest and Blood Tax than they did, and differences under about 0.15 places are noise.

Protect your HP from round 9, when one loss can cost 20

Late damage is a different game. Rounds 9 to 11 have a base of 5 and a cap of 14. From round 12 the base is 7 and the cap is 20, and once 3 or fewer players are left there is no cap at all. Tier 4 effect cards start appearing from round 9, so the boards you face get much bigger too.

Lone Stars is the odd talent out. It pays +4 for each country with no link, +20 on five scattered southern flags (33 to 53), bans Outposts and shuts out the link talents. Bots placed a respectable 3.09 with it as a first pick but won only 7% of the time. The one Lone Stars bot we followed to a win took Treasury second and played Rwanda, Central African Rep., Eswatini, Botswana and Niger with Inland and Sovereign, 96 with Lone Stars. It still had all 32 HP going into round 12. Then it lost 8 HP in each of rounds 12, 13 and 14 and won on 8 HP. A few more points of score and it would have kept most of that.

The defensive talents are for exactly this stretch. Pavise takes 2 off every loss, to a minimum of 1, and was the best first-pick battle talent in our bot data at 3.07. Phoenix heals 8 HP when you take it (up to 32) and brings you back with 5 HP the first time you hit 0; it placed 2.92 as a second pick. Last Stand adds +12 to your total at 10 HP or less, which is how the Blood Tax bot from round 1 got over the line.

When you sit down for the round 7 pick, decide it from the board in front of you and the gold you have planned. Blood Tax then Interest at round 7 is the biggest gold line we found, 66 buys and 210 gold. War Chest then Interest gives 44. If you have the gold, the effect card builds post covers what to spend it on. If your board has Asian links, take Silk Road. If your flags are scattered with no links, take Lone Stars. If the board is good and simply full, take Sixth Seat.