Rise of the AI - Rallying rigged?

[keyth]

After a bit of a break I've been FoGging again for the last few days, practising against the Rise of the AI mod. I don't know whether it's just a statistical anomaly but, in the last 5 or 6 games, when I've got the AI to ~40% and I am 15% or less, it rallies like you wouldn't believe - in one case, 8 units rallied from routed over two turns and just now, versus Macedonians, rallied from 40% to 28% in a turn. In the same games I have had one unit rally from routed. Weird numbers or AI buff?

[TheGrayMouser]

After a bit of a break I've been FoGging again for the last few days, practising against the Rise of the AI mod. I don't know whether it's just a statistical anomaly but, in the last 5 or 6 games, when I've got the AI to ~40% and I am 15% or less, it rallies like you wouldn't believe - in one case, 8 units rallied from routed over two turns and just now, versus Macedonians, rallied from 40% to 28% in a turn. In the same games I have had one unit rally from routed. Weird numbers or AI buff?

Probably law of averages: Near end game you have a larger # of enemy broken units that COULD rally, so more do...

[rbodleyscott]

Rise of AI does not mod the rallying system. The AI gets no buff to rallying (or anything else) in the vanilla game nor in Rise of AI.

Human nature is such that there is a tendency to see one's opponent (be it AI or human) as getting better luck than oneself, when overall this is not true.

The nature of randomness means that random events tend to be clumped rather than evenly spread. If they were evenly spread they would not be random. People tend to notice runs of enemy good luck when it happens, but not notice their own.

Also, as Mouser says, the more units a side has routing, the more chance there is of some rallying, particularly if they routed before suffering heavy losses. The AI is usually less skilled than a player at making units rout without grinding them down first, so routed player units generally tend to be in worse condition than routed AI units, and hence may have additional negative CT modifiers for losses. They may also be autobroken, which guarantees they won't rally.

[keyth]

Thanks for the replies. Having seen (and been guilty of) confirmation bias I try to be a bit more objective - I was monitoring ‘good’ events as well

You have both hit the nail on the head and my only defence is I was tired and not thinking it through: as the potential subject group increases, so will the potential number of rallies. The fact that it happened repeatedly at around the 40% mark is irrelevant but it was that element that made me question the outcomes of rallies.

Thanks again.