(for summary skip to last paragraph, but I think the theory is important)
In poker there is a common phenomenon that often causes otherwise good players to play poorly called "results oriented thinking". What happens is a player will make a good but risky play, like bluffing with a draw, and it happens to fail a number of times in given evening or week. The next time the player is in that situation they will often just fold and avoid the risk, even though it would have made them money in the long run. At a certain point every professional player realizes that analyzing play based on money won or lost is usually counter productive (without using computer tools) because the human mind is so bad at taking in information where randomness plays a big part that it is inevitable you will end up getting tricked by the randomness into playing badly.
Bringing this closer to Tyrant, imagine that in front of you are ten weighted coins where the odds of getting heads on a flip ranges from 45% to 55% (somewhat analogous to surging). If you flip each coin 10 times and are asked to choose which has the highest chance of getting heads, odds of you picking the best coin are going to be about the same as if you picked blind. Even with 40 flips the odds of finding the best coin are slim.
You might be thinking, "So what if we could easily be wrong about which coin is best? In this case the results are all you have." This brings me to my main point: in Tyrant, just like in poker, the results of a small number of battles is not all we have:
-Evaluate decks is an extremely powerful tool as long as you have some decent deck lists. Ordered mode allows close approximations of surging especially when fudge factors are used to account for the extra strategic options available to a human.
-The experience of other players allows pooled knowledge that can defeat otherwise insurmountable sample size issues.
-Experience built up over days against other factions allows decent sample sizes. More often than not 200 battles against other factions with a given deck is more predictive of future results against a certain faction than 20 battles against that faction. The superior sample size far outweighs the variation caused by the deck makeup of a specific faction.
I believe that if a deck is worth using once it's usually worth using at least fifty times. I can count on one hand the number of times I have switched decks in the middle of a war because a faction had decks that countered mine, and I believe that almost everyone overestimates how well they can gauge the decks of a faction after something like twenty battles (without notes). When Ninja had his contest, I believe a significant factor in my victory was sticking with my SC deck through bad stretches when other players would have switched. I think recently showing the same stubbornness with Ogres has also helped my results.