By observing the games, it is possible to get a conclusion that in every game there is an element of manipulation that may effect on the game result.
It looks like a situation where betwean a good player are selected weak allies. It may suggest that reason for that is to balance the odds, which does not mean that the system is selecting players based on a similar MMR (whatever that could be - I thought MMR is some kind of player skill valuation system but now game after game I have more doubts).
Another issue is that the level of our allies gameplay - it looks like it depends on our skills. It's look like our allies could be some kind of bots that dynamically adjust their skill level to ours.
The last time it is possible also to notice that MOStLY in our team we will be able to see 2 players/allies who will act opposite to the team target. For example, if the result of the game is KNOWN from the very beginning - negative (becouse we may know that we are losing too much already), then these two allies will always vote negatively ignoring the / ff or / remake command (or sometimes to push us forward). It's look a like that some RIOT employees or bots, are making their tasks to fulfill certain assumptions, eg the time of the game specified for them - becouse too short game could means that - if team choosing /ff command just after 15min - player selecting system - does not working good - becouse HOW GOOD IT COULD BE IN CASE OF EQUAL TEAMS? (equal teams surrending just after 15 min? - it does it mean "good balance" I guess... right?)
The same pseudo-allies very skillfully / discreetly also trolling their team. Sometimes they run away from TFs, sometimes they stand still in bush. All this involves getting too much advantage in the team. It looks as if their task was sometimes to balance the result of the game or sometimes lead to failure if the allied team wins too much.
Becouse of that it is impossible to say that matches are created on fair-play basis but only to manipulate the final result - appropriate for the RIOT assumptions.