Consider that if David's winning percentage is 63.6, being weaker is actually an advantage, and Goliath's optimal strategy is to abandon most of your resources in order to become the new weakling.

David's winning percentage is reported as 63.6 when he actually assesses his strengths and weaknesses, and adopts a strategy that plays to his strength and hides his weakness. Goliath's optimal (meta) strategy would be to do the same thing, to use a strategy that uses his strength and avoids his weakness.

On the other hand, the data may indeed be tainted by a defender's advantage as you suggest. If Goliath discovers that his strength is that his army has a lot of soldiers, and his weakness is that supporting so many soldiers in an occupied territory involves vulnerable supply lines, his best strategy might be to not invade other countries.

Imagine a game where there are strategies 1-99, where each strategy is more likely to win, but also takes an increasing level of expertise. Imagine also that Strategy 5 has a special property - it beats any strategy numbered in the 90s, but no other.

I would expect the best competitors in this game to prepare to use strategies 5, 89, and the best strategy in the 90's they can handle, and use a meta strategy of switching between these in response to their opponent's strategy. Such a team would defeat another that focused on a higher 90's strategy at the expense of strategy 89, would not be vulnerable to beginners who only know strategy 5, and would still be competitive when the masters switch to strategy 89.