Ok, now im really puzzled why you out of sudden bumped this thread and for.
You planning to write here some small reports about some races you play or what? Or you have some actual questions about something? From that you currently write im not even sure if you actually had read posts here, as you seems to mix the cases of "always winning ai with some race about or better than default one" and "win ai under X turns with right race".

Yes, i meant its way better to write all the race picks used, not only positive.
Yes, those 3 mentioned points are definitely flaws, they could lead to underperform.

Effectiveness stuff in MoO2

Witnessing recent laming spree by some depressingly prolific moronic creative kid on Strategywiki, who is desperately trying to write down any race combination that include Creative there without even playing them first and being unaware of some most basic stuff bout game (this one is so Yuge, it deserve a full quote:
"Although Lithovore accomplishes this by shifting all food requirements to mineral requirements, allowing Lithovore races to do away with the need for shipping food to other colonies via freighters, and cutting costs in the process, unfortunately, Lithovore alone does little to alleviate the need to dedicate a sizeable fraction of the population towards feeding the rest of the population. Assuming a standard Terran mineral-abundant homeworld start, without any production bonuses to boost mineral production, a Lithovore race would need 1 miner to support every 3 pops. (That's the rough equivalent of being a race with the Aquatic, Unification, or +1 Farming picks alone, all of which are much cheaper in point costs to Lithovore!) This means that Lithovore races would also need a pick that boosts mineral production, though with Creative Lithovores, this would be limited to a rich homeworld start due to how expensive both Creative and Lithovore are, at a point cost of 18!",
and he derive thousands of words about way more complicated things starting from it), i recalled what comparison of races and their overall effectiveness been a problem even for people way better than your usual creative kid , so i decided to write a more full version of it here.

But first (and quite important and hard) stuff to stumble at is definition of Efficiency. In relation to MoO2 races we could, for example, declare that race is effective if it can produce 9 or more RP on the first turn, making every possible race combinations thats are not including -ResearchFeudal (and even some that are) equally effective. While being absurd enough already, we could add more, by even lowering the level to be qualified as effective to 8 or more RP on the first turn, making every possible race combinations equally effective.

Yet claiming that every possible race combination are equally effective will not be really helpful (even if that statement would be actually true, due to a wrongly set benchmark as in example above). As it will mean that, for example, that article on Strategywiki should include not only the creative kid's garbage races, but also every other garbage races that possibly to be created in game, as they all could be deemed Effective if set the benchmark too low. Because #1 in this thread is about benchmarks, and also because #6 is about kids, who are rephrasing old Cybersaber's guide instead of writing something of their own (same chow for years and years and years), and the best part of Strategywiki is essentially those kids rephrasings of it (sadly they also do write something on thier own there too), i decided to put it to the same thread, as it about AI playing mostly still (while its actually about PvP aswell too, ofc).

So while we could compare races by any arbitrarily set parameters, better to do it for important ones. If we need to find effective race that is of interest to us, we need to see how good it perform against its goal. It could grow slightly further complicated, as 3 different main strategies of MoO2 (lets call them Blitz, Breed, and Anti-Blitz) have different goals, making comparison between groups somewhat difficult. Say, there is no point to compare production output of some Breed race to one of some Blitz race at say, T100, if the goal of that Blitz race is to finish game at T40. Comparison of production output there wouldnt help either. Easiest is to compare effectiveness inside every main strategy group prior, as its where people do tend to encounter problems already. For simplicity id propose to see a
defining goal to compare per strategy type at given setting as: for Blitz - inability of all opponents to win against the player at most possibly early turn and for Breed - some Combined Production Output at a turn arbitrarily set as a goal (without getting deep on how it exactly calculated and a process behind setting that arbitrary checkout turn; its still pretty easy to grasp as overall empire's total of population, production, research etc). For Anti-Blitz the goal would be some combination of both (and it better to just leave it here, as it exist only in PvP, and this is mostly for beginners, so mostly about AI).

Complicating it further, MOO2 have different Starting Game Settings, Technology Level, Map Size and Tactical/Strategic combat are from most important here. Some race effective under some Settings could perform sub-par at other ones. Its pointlessly and misleading to compare races without setting the actual Settings they're perform at (again, more about it at #3).

But, luckily enough, if we will convert stuff of interest to us under some given circumstances into Imaginary Abstract Units Of Effectiveness Yield (further IAUOEY), we could cover all the cases at once, as being derived from the Michelson-Schwartz's law, and could be easily depicted using a graph.

Graph below shows only approximations of types of MoO2 races in relation to their IAUOEY over Quality of Environment (Map) they're perform at, from worst possible to best possible. Time to mention that Map quality is not entirely random, and depends of Galaxy Generation rules used. Especially difficult stuff is with "worst" Map. For best peak performances there are records set, and such circumstances are overall of more interest for players. Other extreme is way less known (when bad map is encountered, many people tend to go cry for restart), so data about it on shown on graph more theoretical (in short, its not so easy to know for player what Worst possible Map is; and while the same is correct for Best possible Map aswell, there player can refer for collected data for "Best currently known" instead). For simplicity AI playing level is shown just as the X line, despite it, ofc, also depends on Map. Race {examples} are purely for an archetypal reference and based on classix vanilla game and not necessary directly correspond to illustrative only graph.

I dont know how to insert pictures directly to message body here, so ill just put some more direct link to it also, in addition to attachment one below. Afterall the whole point of it to excel with my mad skillz in drawing of graphs.

Comments about Figure's depicting.

Subtype 1, Reliable.
A0: Sigmoid pillar of reliability, maximizing performance on an average expected Map, without dropping performance much in unfavourable extremes {UniTol+ProdLhwRep-SD-GC}.
B1: Logarithmic gambley approach, maximizing performance on a better than average Map, without dropping performance much in unfavourable extremes {UniAqua+Grow+ProdLRhwRep-SD-GC}.
B2: Logarithmic gambley approach, maximizing performance on a worse than average Map, without dropping performance much in unfavorable extremes {dunno, but theory predict its existence}.

Actually the most interesting subtype. Their Area Under Curve is statistically significantly outperform examples from other subtypes, yet they could perform worse than some other race in extremes. Relation between those races are creating metagame, as some of them could be better in AUC total, but strictly worse in performance in some different Map qualities relative to other Subtype 1 races etc. Sadly enough, their relation is also way more complex stuff than relatively simple common misconception we're bout to fix here, so its out of scope of (already too long) article.
Those races are what people usually supposedly to mention as "Effective". Problem is that some of those mentioned are more than often belongs to other subtype, Subtype 3, and some not mentioned, while being effective are belongs to Subtype 2. Even more problem is that its not easy to set a needed tolerance level to definitely demark subtypes. Assume some supposedly B1 race strictly outperform some A0 race at 5% of Maps while slightly losing elsewhere. It seems reliable enough or not? What if its only 1% of Maps, but not only some special peak Map yet etc.

Subtype 2, Peak.
C1: Logarithmic extreme, maximizing performance for the peak on best possible Map, but severely dropping performance in other ones {Feu+2ProdTeleTransdLRhw-Spy-Sci}
C2: Paranoid extreme, maximizing performance for the peak on worst possible Map, but severely underperform in other ones {DemoSub+Res+LAhwRep-SA-GC}

Those race are best under special conditions only. A race races to set records, not some reliable ones to play with something at stake. C1 is easier to get as there is a good video of https://www.youtube.com/watch?v=C0zSm1x_sTs]Rocco playing my record race (it would been even better if he mentioned there about the way more impressive fact that im played it in 18 turns in the real (no saveload, unknown map) run first, then improved it to16 turns playing from the save, and after that provided it to him to record a video. He still chow those 16 turns, reluctantly mentioning bout "save was given to me"). C2 is theoretical, as it intended to be played on "worst" map, not possible in actual vanilla map generator (no other satellite inside of Thorium range to land on, so playing from one planet only), but can be played with rules forbidding settling of satellites or capturing other colonies on any map. While Subtype 2 races generally statistically perform even worse than kids races of Subtype 3, unlike them they are Effective still, as they do actually reach the worthy goal they have to reach (outperform every other race is definitely a worthy goal).
Strictly speaking, Peak is also could be seen as B1 of Reliables taking to extreme. Under some settings they even could entirely overlap: say in Prewarp the Ultimate anti-AI Teleblitz (UniTeleTransdLRhwLg-GC-Spy), not only the most Reliable race, but also the Peak one (even on best possible Map FeuBlitz cannot outperform it due to need to actually do some research first); while in Average the C1 example race is the Peak one (but not Reliable one, that is still Teleblitz). And on Advanced neither of them are Reliable or Peak, the Settings matter that much.

Subtype 3, Junk.
D1: Just abstract representation of some mediocre race, that is significantly underperforming in comparison to Reliable, and never reaching Peak, while obviously doing strictly better than AI {something various creative kids and other lamers call "builds", there are shitload of them in innternets}.
D2: Just abstract representation of some other mediocre race, that is significantly underperforming in comparison to Reliable, and never reaching Peak, while obviously doing strictly better than AI {something various creative kids and other lamers call "builds", there are shitload of them in innternets}.

And here are those races those doesnt reach the goal they need to, or trying to reach not worthy one. For example, as mentioned in #1, there is no weapon player need above nukes to win any vanilla AI, so if race is set (and even reach) a goal of researching some other shining tech of victory to just slowly win with it, the goal itself is not worthy (unlike the goal of Peak races), even if race is actually succesful in it. Thats the most common mistake of new/lame players in a nutshell; they tend to not know actual benchmarks of race's performance, so often mistook weak performance of own "builds" as good enough (as it still can win AI (and we already know from #1 that there is no actual worthiness in it, its given)). Its easy to see that Junk races are created in hope of creating Reliable one, but failed attempt, either due to inability to keeping the pace with other, better races; or, due to player's ignorance, setting the goal too low (to bother with results). There even added some small intervals for them to outperform some Reliable races on it (as mentioned above, strictly speaking there could be eventually some uncertainty between Bx & Dx, depending on outperforming interval and overall AUC), but small enough to be considered significant.

Bonus stuff, Antiraces.
E0: Race intentionally designed so bad, so even winning over AI is possible as some peak performance on a very favorable Map {-Food-Prod-SciCyber}

While there indeed are races that are intended to perform worse than AI, and thats their goal. Strictly speaking they are close to Peak races some twisty but similar way, just goal of Antiraces is severely scaled down in comparison, from outstanding performance to merely winning.

So in summary: there are not so many worthy Effective races to talk about per Setting: only Reliable and Peak. Most common mistake is to present some Junk race as Reliable, due to overvaluing of its (actually weaker than supposed) performance.
But, on the other hand: there are many Settings to compare at; and not every race worthy in one is also worthy in other. Most common mistake is not even realizing the difference between settings, like some typical creative kid who write list of "effective creative builds", and while Creative indeed would be a staple for a AdvancedTech race (while still very different one than that kid will usually write about), the kid will ensure its not something he have in his mind, as you would be able to immediately read his bright thoughts about importance to get an Automated Factories and Soil Enrichment early for this race (so definitely not AdvancedTech is meant (curious stuff is what is actually meant there tho, as it could looks author providing shiney advice to get Autos and Soils instead of rushing to research some PhasingCloak straight from the turn 0 (in a turn such behavior is a common cause of AI failures in this games, as they often researching some Phasors without getting into Autos at some horribly late T200))).

So dont fall into trap of calling some weaker for a given conditions race as "effective" even if you "invented" it (likely no such thing happened in this millennia), if it cannot outdo other races in reliability or peak performance. The mere fact it can outdo some other races or AI is not justifying it. Also dont fall into trap of not mentioning of Settings, they are really do matter.

Updated YT video description.