Robot Wars

• Chestnut is the basis for the other robots. Each nut has his own key strategy, but if that is not triggered, they fall back to Chestnut's strategy. The way Chestnut works is that he computes the profit (cost vs income) for building a factory and then a hotel in every city on the board, which means he does 18 calculations. He then does the most profitable action that he can do. He then does it all over again to see if he should buy a ticket for his next turn. That's about it except for some special tweaks, such as generally building two hotels if he visits an empty city. With this algorithm, I did not know what Chestnut would eventually do or how successful he would be. In the beginning he killed everybody, and we humans had to adjust our strategies. I tweaked him a little, but he has primarily stayed the same, calculating the best move turn after turn.

• Cashew and Almond are factory lovers. Cashew will build a factory if a certain number of hotels appear in a city, and he adds a bonus for factory building when he falls back to Chestnut's scoring. Almond derives from Cashew, but he is even more aggressive. For example, if he lands in a city where someone else is, he immediately buys a factory without even looking at the hotels. He will also try harder to ensure his factories get the "longest chain" bonus.

• Walnut is not much different from Chestnut. She will start the game taking out two small loans over two turns. She will also build factories in the first cities she visits with an open first-factory slot.

• Pecan looks ahead a bit further than the others. He tends to stay longer in cities to ensure that he builds enough hotels to score. He will also build the first factory in a city if another player has built hotels there.

• Peanut is an odd nut. He is the only one who will venture out occasionally without a loan. As the game goes on, he tends to disregard the cost of tickets in his pursuit for pure income.

Nut	Games Played	Total $ Earned	Min $ Game	Max $ Game	Aver $ Per Game
Almond	6777	1000709	28	339	147
Cashew	6735	963961	15	315	143
Walnut	6784	920071	32	292	135
Pecan	6824	891897	10	318	130
Chestnut	6559	853094	2	316	130
Peanut	6305	709331	-2	287	112

Hail to Almond. It is interesting that even the best nuts occasionally have a dreadful game. I then started tweaking each nut, running the contest again, and making adjustments to see if I could raise a its average income per game. It occurred to me too that some useful human strategies could pop up out of this. So here they are -- nut tips.

Nut Tips to You Humans

• The number one tip the nuts tell us is one word: factories. Your goal should be to have most of your factories on board by the end of the second session. This does not mean you should run around dropping factories willy-nilly. If you build a factory in an empty city, it will probably be just you who does all the hotel building. However, if others have built some hotels in a city and the first slot is empty, jump on it. If you plan to stay awhile to build hotels, build the factory first. Although a factory does not break even until four hotels are in the city, remember that factories pay out every session, so build early and often.

• Should I try to get in as many cities as possible to get that bonus? The nuts say no. Early on, I built some robots whose primary strategy was to fly around and build hotels everywhere. It was such a dismal failure that I discarded them and started over.

  However, do not ignore the bonus entirely. It just should not be your primary focus. I have had problems with the nuts when occasionally they refuse to fly anywhere because cheap tickets are unavailable, and they end up building six or more hotels in one city. This is a degenerate case, and the disease always causes the nuts to perform terribly. I eventually had to write special code to avoid the problem.

• Does any seat have an advantage? Is it best to go first? Last? This is not a tip but an observation. The nuts say no. I watched this for probably a 100,000 games and saw no distinct advantage to being in any seat. For awhile I thought being last might have a minor advantage, but this did not hold up.

• How much money should I borrow before I start? At one point I was testing with four Chestnuts, and when I changed from one big loan to two small loans over two turns, Chestnut got a big boost in income, almost $6 per game. That was shocking. Of course, you need to be careful here because all four players were delaying two turns. The big increase did not hold up in normal games, but it did not hurt. You get more money, and the penalty and payoff is easier come payday. I would need to do more testing here to check this out, but the results suggest that taking out two small loans instead of one big one may be better. I've adopted this strategy in my games and immediately hit a winning streak. However, there is not enough evidence to say it is better, but there is enough to say it certainly does not hurt.

  As for not borrowing any money, this was not a good strategy if done unconditionally every game. However, it may be a good idea depending on the initial ticket setup. If there is a $4 ticket available with a $1 connector to a cheap city with maybe some $1 Cairo tickets, then it may be a good idea to jump out loan-less. You get started cheap and you steal some great tickets from someone else. Note that this is all personal observation, and it is something I could not test with the nut tournament.

• Should I pay off my loans or should I extend them? I first had the nuts hold only $9 aside before paying off their loans. I then increased it to $20. Big difference. It is clearly much better to not to pay off loans if it means you would have less than $20 on hand. I then tried it with $30, but it did not make it better. The rule thumb is you should set aside at least $20 before you decide to pay off loans. You goal is to not return for another loan in the next session.

• How many hotels should I build in a city before flying on? We are talking here about when you are a city's first visitor. It is not a good idea to build just one hotel and then fly on. This was part of the "fly-alot-build-everywhere" robot strategy that I discarded. If your only hotel in a city is the first one, then it won't be long before you won't have any. It is better to build two hotels, first for your protections, and second because it sets you up better to build that first factory. I did run some extensive tests with robots that forced them to build three hotels, but this had no effect on the outcome. The current robots all build two hotels when they are a city's first visitor.

• What should I do about buying tickets ahead of time? At one time Chestnut would buy tickets because he thought he needed them or because they were just cheap. I even had one nut buy every $1 ticket available. None of these strategies were particularly profitable. After extensive testing of all kinds of ideas, what it finally came down to was this: If you want to go somewhere next turn or so and there is a $1 or $2 ticket available, but it. Otherwise, save your money.