> First match a Draw. After that fruit wins, after that it lost.

The sample size is too small, e.g. Fritz 5.12 could beat Rybka 3 three times in a row, then go and lose 97 games/matches in a row, that's why you need bigger sample sizes.

> Im talking about Toga 1.4 the Free Engine.

The rating lists already produce big sample sizes that are relatively more reliable that what you can do at home (unless you play more games than them), please check:

CEGT Rating list

CCRL Rating list

Both lists agree that Toga is stronger than Fruit (even than strongest private fruit.)

> did u mean the number of games weren't enough to see who's stronger?

Yes. For certainty thousands of games are needed. You can download software like Bayeselo, input the games from the engines, and you'll get estimated ratings, i.e.

Engine 1: 2800 +50 -50
Engine 2: 2700 +75 -74

The plus and minus info means that the rating is inside that range, so Engine's 1 rating is between 2750 and 2950, and Engine 2's rating is between 2626 and 2775. And you can see these ranges overlap, so you still can't be sure who is stronger. Once you play enough games that these ranges don't overlap, I think you've achieved a 97% certainty.

> So even the Toga || 1.4 is stronger than any Fruit?

Toga II 1.4.2 JD 4CPU specifically seems stronger, but you can see ratings overlap so Private Fruit could be stronger (though you can't have private Fruit anyway.)