CONQUEST Bots: What Easy, Hard and God Mode Really Do
Easy bots won 5% of mixed games, Hard and God Mode 36% each. What each CONQUEST bot level attacks with, when it gives up, and how to beat it.

Each CONQUEST bot level is one setting turned up: how many armies per defender the bot wants before it attacks. Easy wants 2.2, Medium 1.6, Hard 1.3 and God Mode 1.05.
In 1,000 four-way World map games with one bot of each level, Easy won 5%, Medium 24%, Hard 36% and God Mode 36%. The top two are the same bot with a lower bar for the first 70 rounds, not a different mind.
CONQUEST is a RISK-style game and is not affiliated with Hasbro, the maker of RISK. The numbers below describe CONQUEST and its bots.
Four levels are four thresholds and a few switches
The bot code lives in shared/bots.js. Every level shares one decision routine and reads its own profile.
| Level | Armies per defender to attack | Attacks per turn, at most | Fortify chance | Plans a region | Hunts weak players |
|---|---|---|---|---|---|
| Easy | 2.2 | 2 | 40% | No | No |
| Medium | 1.6 | 6 | 80% | Yes | No |
| Hard | 1.3 | 14 | 100% | Yes | Yes |
| God Mode | 1.05 | 99 | 100% | Yes | Yes |
The ratio is the attacking tile's armies minus 1 (one must stay home), divided by the defenders. A bot attacks the best target that clears it.
A ratio of 1.0 is parity. Jason Osborne's Markov chain analysis of RISK finds that with equal armies on both sides the attacker takes the territory more than half the time once each side has at least 5. Our exact figure for 6 armies on the tile against 5 defenders is 50.6%. The table further down shows what each level's threshold buys.
Two details matter. Easy bots place armies on a random border tile and pick draft territories at random, while the other three score regions and move toward the bonus they can finish. Easy also holds its cards until it has five, or until the next set is worth 10 armies or more. The others trade as soon as they have a set.
Hunting means the bot picks the weakest opponent and, once that player is down to four territories or fewer, adds a bonus for finishing them. The bonus grows with the cards that player holds, and an elimination hands those cards to the bot that made it.
The bar falls every round, and the levels meet in the middle
A round is one pass of every living player. The threshold is not fixed. In the code it is the profile ratio minus the round number divided by 250, with a floor of 1.02. Every bot gets bolder as a game drags on.
| Round | Easy | Medium | Hard | God Mode |
|---|---|---|---|---|
| 1 | 2.20 | 1.60 | 1.30 | 1.05 |
| 25 | 2.10 | 1.50 | 1.20 | 1.02 |
| 50 | 2.00 | 1.40 | 1.10 | 1.02 |
| 100 | 1.80 | 1.20 | 1.02 | 1.02 |
God Mode is on the floor by round 8. Hard gets there at round 70, Medium at 145 and Easy at 295. After round 300 the floor drops to 0.9 and the cap on attacks per turn disappears: the code calls this stalemate pressure.
That is why the gap between Hard and God Mode is smaller than the labels suggest. A God bot is only more eager in the opening, and the opening is where a human player is also busy claiming a region.
The smallest stack that will attack you
Here is the same rule turned around, as the tile a bot needs before it hits yours at round 1. The percentage is the exact chance its blitz takes the territory under Wild Dice, from listing every roll and working the battle through. The dice odds post shows the method.
| Defenders on your tile | Easy | Medium | Hard | God Mode |
|---|---|---|---|---|
| 1 | 4 (92%) | 3 (75%) | 3 (75%) | 3 (75%) |
| 2 | 6 (89%) | 5 (79%) | 4 (66%) | 4 (66%) |
| 3 | 8 (91%) | 6 (77%) | 5 (64%) | 5 (64%) |
| 4 | 10 (93%) | 8 (83%) | 7 (74%) | 6 (64%) |
| 6 | 15 (97%) | 11 (86%) | 9 (73%) | 8 (64%) |
| 8 | 19 (98%) | 14 (89%) | 12 (79%) | 10 (65%) |
| 12 | 28 (99%) | 21 (94%) | 17 (83%) | 14 (66%) |
Read the first number as armies on the attacking tile. A Hard bot needs 5 armies on a tile to attack your 3. A God bot needs the same 5 and will take it 64% of the time.
The percentages are the useful part. A God bot attacks at odds that are not much better than two in three, and bots always blitz: they roll until they capture or until one army is left on the tile. At the minimum stack, roughly one God attack in three therefore ends with the attacking stack gone and a single army sitting on that border.
That is the opening for a counter-attack. The dice rules in the Hasbro rulebook are the ones the bot is gambling on: the attacker rolls up to 3 dice, the defender up to 2, and a tie goes to the defender.

Hard and God Mode win equally often
Method: the real rules engine and the real bot brains, World map, Wild Dice, auto-deal. Seats were rotated through all four positions so no level always moved first, 250 games per rotation.
| Four-way table, one bot of each level | Wins in 1,000 games |
|---|---|
| Easy | 5% |
| Medium | 24% |
| Hard | 36% |
| God Mode | 36% |
| Two players, 250 games per seat order | Stronger level wins |
|---|---|
| Medium against Easy | 500 of 500 |
| Hard against Easy | 500 of 500 |
| God Mode against Easy | 500 of 500 |
| Hard against Medium | 79% |
| God Mode against Medium | 77% |
| God Mode against Hard | 52% (257 of 499) |
At 499 games, a 52% result has a margin of about four points either way, so we read God against Hard as a coin flip. Bots use random numbers, so a re-run moves each percentage by a few points.
Seat order is the reason both seat orders were played. With two players and equal bots, the first seat wins about three games in four, as the first-player post measured.
The profile table shows why it comes out level. Both levels plan, hunt and fortify every turn. God Mode's cap of 99 attacks sounds large, but a God bot averaged 6.3 attack starts per turn against 6.2 for Hard, so the cap of 14 is not what separates them.
Easy does not lose slowly, it just does not finish
With four bots of one level, the median game took 306 rounds for Easy, 137 for Medium, 35 for Hard and 30 for God Mode. Easy's median sits just past round 300, which is where its own threshold finally reaches the floor and where the stalemate rule switches on.
An all-Easy table ends when the thresholds have fallen far enough for the bots to start fighting. That suits a beginner who wants a long, forgiving game and is a poor way to practice for a hard table. The real-time versus turn-based comparison quotes 27 rounds for Hard bots from 40 games. This run used 250 and measured 35, so treat both as the same ballpark.
Alliances: the higher the level, the colder the answer
When you propose an alliance to a bot, the server rolls an accept chance by level: 90% for Easy, 65% for Medium, 40% for Hard and 20% for God Mode. The answer comes 1.2 to 2.7 seconds later.
CONQUEST does not enforce alliances, so what matters is what the bot does afterward. Easy, Medium and Hard bots never attack an ally. A God bot filters allies out too, unless nothing else is worth hitting, and then it takes the ally. The betrayal is announced to everyone.
Our rule of thumb is to trust an alliance with a bot only while it has another target. Cards are the usual reason to land at the top of a hunter's list, and when to trade in RISK cards has the set values.
What to do against each level
These are recommendations from the numbers above, not a ranking of anyone's skill.
- Against Easy: your border can be thin. It needs 8 armies on a tile to attack 3 defenders, and it only starts two attacks a turn. Use it to learn the cards and the region bonuses, as in which continent to take first.
- Against Medium: it targets the region with the best bonus for the fewest armies to clear. Hold the region that looks cheapest to it, and watch that one first.
- Against Hard and God Mode: keep each border tile above the table's numbers. It picks the target with the best ratio, so the thinnest tile next to its biggest stack goes first.
- After a blitz fails: the attacking tile is left with one army. Hit it that turn.
- When you are small: once a player is down to four territories, a Hard or God bot goes after the weakest one first, and more cards make that player a better target.
You can set the level in the lobby of CONQUEST, and the options are explained on the how to play page. Medium is the default.
More from the blog.

RISK-Style Browser Games: Which Need No Account?
Five RISK-style games you can play online, checked on their own sites: who asks for an account, who has bots, and how long a 2-player game runs.
Oct 18, 2026 · 5 min
CONQUEST Capitals Mode: 4 Players Finish in 9 Rounds
Capitals mode ended 4-player World games in 9 rounds instead of 32. How the capital is picked, why it starts exposed, and how to play for it.
Oct 16, 2026 · 6 min
Which Continent to Take First in CONQUEST (World Map)
Oceania and South America each pay 2 armies a turn for about 9 expected losses. Oceania has one door. All six regions of the World map, computed.
Oct 08, 2026 · 6 min