If genetic fitness actually matters, is life limit still necessary?

jasonrohrer · 2019-10-30 19:27:22

Here's K=160

Baby death rate 0.1, Lowest score seen = 50.455030, highest = 57.137137
                     Overall Ave = 54.101095

Baby death rate 0.2, Lowest score seen = 43.166256, highest = 52.814564
                     Overall Ave = 48.036113

Baby death rate 0.3, Lowest score seen = 35.835678, highest = 49.667531
                     Overall Ave = 41.807108

Baby death rate 0.4, Lowest score seen = 30.001798, highest = 42.017283
                     Overall Ave = 36.081239

Baby death rate 0.5, Lowest score seen = 23.760806, highest = 35.551679
                     Overall Ave = 30.195679

Baby death rate 0.6, Lowest score seen = 18.001891, highest = 30.618424
                     Overall Ave = 24.094430

Baby death rate 0.7, Lowest score seen = 13.020215, highest = 25.153200
                     Overall Ave = 17.923098

Baby death rate 0.8, Lowest score seen = 7.687261, highest = 17.531027
                     Overall Ave = 11.919100

Baby death rate 0.9, Lowest score seen = 2.740958, highest = 9.610792
                     Overall Ave = 6.063054

jasonrohrer · 2019-10-30 19:30:24

Once we get up to K = 320, we get some nice banding in the highest/lowest, where the 0.1 player is always better than the 0.2 player, etc:

Baby death rate 0.1, Lowest score seen = 51.012993, highest = 55.964396
                     Overall Ave = 54.028521

Baby death rate 0.2, Lowest score seen = 44.206341, highest = 50.923926
                     Overall Ave = 47.852101

Baby death rate 0.3, Lowest score seen = 38.548593, highest = 45.778896
                     Overall Ave = 42.096366

Baby death rate 0.4, Lowest score seen = 31.807311, highest = 40.935173
                     Overall Ave = 36.074301

Baby death rate 0.5, Lowest score seen = 26.242167, highest = 33.881332
                     Overall Ave = 30.019951

Baby death rate 0.6, Lowest score seen = 20.482026, highest = 27.571718
                     Overall Ave = 24.044021

Baby death rate 0.7, Lowest score seen = 14.112724, highest = 21.512703
                     Overall Ave = 18.032687

Baby death rate 0.8, Lowest score seen = 8.794995, highest = 15.130280
                     Overall Ave = 11.804254

Baby death rate 0.9, Lowest score seen = 3.805938, highest = 8.358687
                     Overall Ave = 5.970621

miskas · 2019-10-30 19:32:37

"How much your score jumps after an offspring dies in OHOL depends on how surprising the result is."

So we have the player A that constantly dies at 60 no matter what his mother is, gets born to a player B that his bbs on average die at 20.
What is the unexpected event The player A die at 60 again or at 20?
Player A dies at 20 is the unexpected event cause he is supposed to die at 60 no matter what his mother is.

When you count Ranking you have to consider the Ranking of the opponent, The ranking of player A is 60 and he died at 20 with this mother so something has to be gone wrong. but the player B with the current system goes unaffected by this unexpected event.

Score gain has to calculate the Average age of death of the bb to see if this mother has a positive or a negative influence.
Score gain= Age of death of kid - the Average age of death of the kid.

------------the opposite example:----------
we have the player A that constantly dies at 20 no matter what his mother is, gets born to a player B that his bbs on average die at 50.
What is the unexpected event The player A die at 20 again or at 50?
Player A dies at 50 is the unexpected event cause he is supposed to die at 20 no matter what his mother is.

When you count Ranking you have to consider the Ranking of the opponent, The ranking of player A is 20 and he died at 50 with this mother so something has to be gone Right. But the player B with the current system goes unaffected by this unexpected event.

Last edited by miskas (2019-10-30 19:41:37)

jasonrohrer · 2019-10-30 19:35:10

However, the problem with K=320 is that the climb is SUPER slow. This is for 50,000 babies after a 50,000 warmup.

Here's K=320 for 100 babies after a 100-baby warm-up:

Baby death rate 0.1, Lowest score seen = 13.955712, highest = 23.907755
                     Overall Ave = 19.392539

Baby death rate 0.2, Lowest score seen = 12.736786, highest = 21.831406
                     Overall Ave = 17.915530

Baby death rate 0.3, Lowest score seen = 11.487394, highest = 21.034583
                     Overall Ave = 16.459338

Baby death rate 0.4, Lowest score seen = 9.975432, highest = 17.030706
                     Overall Ave = 13.093496

Baby death rate 0.5, Lowest score seen = 7.399349, highest = 14.899073
                     Overall Ave = 11.247081

Baby death rate 0.6, Lowest score seen = 6.939977, highest = 10.980586
                     Overall Ave = 8.787001

Baby death rate 0.7, Lowest score seen = 5.871116, highest = 10.516907
                     Overall Ave = 8.445484

Baby death rate 0.8, Lowest score seen = 3.037865, highest = 5.947009
                     Overall Ave = 4.480372

Baby death rate 0.9, Lowest score seen = 1.974833, highest = 3.439807
                     Overall Ave = 2.728647

Note that we're nowhere near our desired averages after 100 babies. We're climbing SO slow...

Going back to K=10, for the same 100 BB run and warm-up, we see that we converge on our averages:

Baby death rate 0.1, Lowest score seen = 48.196386, highest = 59.152621
                     Overall Ave = 55.091739

Baby death rate 0.2, Lowest score seen = 37.621609, highest = 57.094743
                     Overall Ave = 46.423213

Baby death rate 0.3, Lowest score seen = 25.830947, highest = 51.776318
                     Overall Ave = 41.564818

Baby death rate 0.4, Lowest score seen = 26.290739, highest = 47.248168
                     Overall Ave = 36.801265

Baby death rate 0.5, Lowest score seen = 10.708538, highest = 42.898432
                     Overall Ave = 26.807254

Baby death rate 0.6, Lowest score seen = 8.032299, highest = 35.318761
                     Overall Ave = 21.572470

Baby death rate 0.7, Lowest score seen = 6.288324, highest = 34.911071
                     Overall Ave = 17.946457

Baby death rate 0.8, Lowest score seen = 5.902460, highest = 27.932349
                     Overall Ave = 12.562949

Baby death rate 0.9, Lowest score seen = 1.227528, highest = 15.066245
                     Overall Ave = 6.729200

Maybe looking at 50,000 babies was misleading a bit. For 100 babies, we still have huge overlaps in these bands, which isn't great. But maybe K doesn't need to go up to 320. What K balances overlap with accurate averages over 100 babies?

Saolin · 2019-10-30 19:40:42

miskas wrote:

"How much your score jumps after an offspring dies in OHOL depends on how surprising the result is."
So we have the player A that constantly dies at 60 no matter what his mother is, gets born to a player B that his bbs on average die at 20.
What is the unexpected event The player A die at 60 again or at 20?
Player A dies at 20 is the unexpected event cause he is supposed to die at 60 no mother what his mother is.
When you count Ranking you have to consider the Ranking of the opponent, The ranking of player A is 60 and he died at 20 with this mother so something has to be gone wrong. but the player B with the current system goes unaffected by this unexpected event.
Score gain has to calculate the Average age of death of the bb to see if this mother has a positive or a negative influence.
Score gain= Age of death of kid - the Average age of death of the bb.

Yeah! I was just working on posting an idea about this. What if gene score compared how long a player usually lives for to how long they lived in the current life and awarded a score based on that? So having say a new player that tends to die early live until 60 would give their family a big boost, but that experienced kid you had who always lives to 60 gives everyone a smaller boost when they do so since it's more "expected".

I think it might remove the upper and lower score limit though.

Last edited by Saolin (2019-10-30 19:45:08)

jcwilk · 2019-10-30 19:45:38

jasonrohrer wrote:

Maybe looking at 50,000 babies was misleading a bit. For 100 babies, we still have huge overlaps in these bands, which isn't great. But maybe K doesn't need to go up to 320. What K balances overlap with accurate averages over 100 babies?

Well let's take a step away from the math for a second and answer some questions, even if they're not 100% perfect answers (since it's hard to know what the perfect answer is without seeing the big picture) it might lead us to a good-enough algorithm:

How many successful (let's step away from 60 year old for a moment since in another thread it sounds like you were getting on board with a lower age being considered successful, 40 in that case but others were leaning towards 20, so let's just say
"successful" for now) children should a new player need to bear in a row without any failures before they make their way up to being correctly rated in the 90% percentile? (putting aside what being rated in the 90th percentile looks like for now)

How many failures in a row should an established 90th percentile player endure before their score dips to match that of a correctly rated 80th percentile player?

If an established 80th percentile player suddenly turned into a 90th percentile player, how many successes in a row should they need before they were rated as a 90th percentile player?

Should "established" be a flat state once you reach it or should one continuously become more and more established and rigidly ranked?

How many births should be required before someone is fully "established", or in the case of continuously progressively established, how many births should be required before it "feels" like it's slowed down significantly?

If we can nail down some line-in-the-sand answers to those questions then the process of finding sliding K values to fit should be trivial.

edit: adjusted language to be "should" to make it clearer

Last edited by jcwilk (2019-10-30 19:48:57)

Kaveh · 2019-10-30 19:47:43

jasonrohrer wrote:

Maybe looking at 50,000 babies was misleading a bit. For 100 babies, we still have huge overlaps in these bands, which isn't great. But maybe K doesn't need to go up to 320. What K balances overlap with accurate averages over 100 babies?

this

jcwilk wrote:

Agree, sounds good. Bumping up the K value and then taking a look at the first few dozen could be useful too (edit: ie, before they've settled since that's the biggest downside of a high K value, unless you want to gradually raise the K value the more they play which is I think how chess usually does it)

Gradual increase of K value w/ the # of lives played (if you have that data)?

olooopo · 2019-10-30 19:48:09

What happens if you make the constant K a function dependent on certain individual factors (e.g. number of total lives, number of bbs, current bb death rate)?

Last edited by olooopo (2019-10-30 19:48:38)

jasonrohrer · 2019-10-30 19:54:07

The problem with talking about percentiles is that it depends on how good the other players are, right?

And yes, I suppose an event would be unexpected depending on how long a BB is expected to survive, so the mom must have helped, etc. The problem with this is that it's opaque and hard for players to reason about. They look at the chart, see a BB that lived to 60, and wonder why they got almost no score boost. They don't know that BB was a pro player who always lives to 60, right? The Elo is very transparent. A given baby in a given situation always raises the score by the same amount. So players can get used to this, and the numbers don't defy expectation or require additional explanation.

I'm now testing different K values over 100 babies (with 100 warmup) to see how far the average is from our target and how much overlap there is between bands. Here's K = 10:

Baby death rate 0.1, Lowest score seen = 43.886356, highest = 58.759254
                     Overall Ave = 52.753550
                     Desired Ave = 54.000000
                     Ave error = 1.246450

Baby death rate 0.2, Lowest score seen = 33.106839, highest = 56.825644
                     Overall Ave = 45.707789
                     Desired Ave = 48.000000
                     Ave error = 2.292211

Baby death rate 0.3, Lowest score seen = 27.029225, highest = 54.212691
                     Overall Ave = 42.926473
                     Desired Ave = 42.000000
                     Ave error = 0.926473

Baby death rate 0.4, Lowest score seen = 20.566258, highest = 44.684065
                     Overall Ave = 32.757150
                     Desired Ave = 36.000000
                     Ave error = 3.242850

Baby death rate 0.5, Lowest score seen = 14.084637, highest = 40.417921
                     Overall Ave = 30.001167
                     Desired Ave = 30.000000
                     Ave error = 0.001167

Baby death rate 0.6, Lowest score seen = 11.799152, highest = 37.459471
                     Overall Ave = 25.693734
                     Desired Ave = 24.000000
                     Ave error = 1.693734

Baby death rate 0.7, Lowest score seen = 3.819207, highest = 31.357374
                     Overall Ave = 14.958863
                     Desired Ave = 18.000000
                     Ave error = 3.041137

Baby death rate 0.8, Lowest score seen = 3.613492, highest = 25.487554
                     Overall Ave = 12.232301
                     Desired Ave = 12.000000
                     Ave error = 0.232301

Baby death rate 0.9, Lowest score seen = 0.910338, highest = 12.727012
                     Overall Ave = 5.457854
                     Desired Ave = 6.000000
                     Ave error = 0.542146

K=10.000000, Total ave error = 13.218468, Total overlap = 292.583986

jasonrohrer · 2019-10-30 20:01:34

Running for a bunch of K, I get this surprising result:

K=10.000000, Total ave error = 24.329760, Total overlap = 238.688103
K=20.000000, Total ave error = 17.059244, Total overlap = 121.010260
K=30.000000, Total ave error = 13.280489, Total overlap = 42.708661
K=40.000000, Total ave error = 19.894754, Total overlap = 31.334682
K=50.000000, Total ave error = 16.355584, Total overlap = 21.686105
K=60.000000, Total ave error = 31.114395, Total overlap = 20.514129
K=70.000000, Total ave error = 33.140592, Total overlap = 20.662166
K=80.000000, Total ave error = 37.771609, Total overlap = 28.252469
K=90.000000, Total ave error = 56.370452, Total overlap = 34.422392
K=100.000000, Total ave error = 65.414879, Total overlap = 47.954260
K=110.000000, Total ave error = 67.483215, Total overlap = 33.138665
K=120.000000, Total ave error = 81.438414, Total overlap = 28.372636
K=130.000000, Total ave error = 86.762028, Total overlap = 40.667230
K=140.000000, Total ave error = 94.551536, Total overlap = 49.340552
K=150.000000, Total ave error = 102.680197, Total overlap = 42.131977
K=160.000000, Total ave error = 111.954525, Total overlap = 57.593071
K=170.000000, Total ave error = 111.793969, Total overlap = 38.609943
K=180.000000, Total ave error = 115.247696, Total overlap = 48.937061
K=190.000000, Total ave error = 125.418291, Total overlap = 45.870941
K=200.000000, Total ave error = 132.552407, Total overlap = 51.118812

Looks like K=50 minimizes both the average error after 100 BB and the overlap between bands.

fug · 2019-10-30 20:04:25

Back to the start of the thread the life limit is clearly harmful in cases where the end is near as babies keep spawning themselves into mothers who won't feed them thus running them out of lives. However, if your meme score is low enough you might as well just kill yourself mid life to get around the tool limit.

For example: You want to help make your town an engine so they've got water and what not. First life learn tongs, firing kiln, hammer, newcomen hammer, roller.

Convert all your iron to steel
Smash one iron for the tank with your hammer
Make your Pistons and rods.
Kill yourself.

Get reborn and learn bore, lathe, knife, oven, free space.

Bore two pistons
Bore three rods
Lathe the stuff
Fetch rubber from wilderness
Combine stuff together
Finish engine.

Kill yourself again and now start actually playing. Unless your gene score is high enough to have 9 slots then killing yourself is probably the better option due to the increase chance of fuck ups when introducing a second person to engine making. Fucking up also means you need +1 tool slot bringing your required slots to 10 which is the max slots atm.

So long story short, score only really matters if you're playing a single life in a single sitting. If you're going to binge and play a few lives it doesn't really matter to have meme score and you'll actively be sinking someone else's score to get around the restrictions of tool slots.

Last edited by fug (2019-10-30 20:16:44)

jcwilk · 2019-10-30 20:05:56

jasonrohrer wrote:

The problem with talking about percentiles is that it depends on how good the other players are, right?

Yeah but that's the whole point isn't it? To differentiate a player who is better than everyone else from everyone else. If we overfocus on "a player who can get 80% of babies to age 60" then all those calculations go right into the trash the minute it's decided that living to 40 is the new living to 60, or if a change to the game made it where reaching age 60 suddenly became super difficult. So if we abstract it a bit and say "there's some kind of behavior that roughly falls under the umbrella of 90th percentile excellence, what should being a 90th percentile player look like in terms of how the rating responds? How about when they up their game and get to 95th percentile?" I'm sure most of us have gotten really good at some online game at some point and experienced the way its rating system worked as you got closer and closer to the peak, and whether that was terrible and frustrating like sisyphus or appropriately scaled and rewarding.

Focusing on that experience first and then looking at the numbers after we have an idea of what things -should- be, even if they're handwavy and difficult to nail down like "90th percentile" means we can say things like "well 90th percentile players should take 20 births to reach their appropriate rank, and this algorithm makes it so that a mother that keeps her kids alive 80% of the time ends up taking 20 births to have their rank settle down. Is 80% survival of children a 90th percentile behavior?" then it can be scaled from there accordingly. 90th percentile could even be translated precisely into a baby survival rate from the current data if you so desired.

Edit: and then if later after changes to the game 90th percentile behavior was actually a different baby survival rate then you just need to scale it to match it up again with the percentiles, or perhaps you could build a mechanism into the game that does this automatically, or perhaps the algorithm would be agnostic to this somehow. How the algorithm works is less important than how well it achieves the goals, though I agree that making it transparent will also be important... But there's more than one way to make something transparent, of course. Having extensive details about your rank, how it changed, and why on the death screen would go a long way I imagine.

Last edited by jcwilk (2019-10-30 20:12:18)

Kinrany · 2019-10-30 20:15:37

As I understand it, genetic fitness is a way to incentivize players to do the correct thing. There's no point in comparing players to each other, right? So I think Elo score is not a good analogy.

Neither are fitness functions from genetic programming. A fitness function measures performance of a single agent, not a gene. That would be similar to measuring performance of a single life in OHOL.
A better analogy for measuring performance over multiple lives would be reward functions from reinforcement learning.

miskas · 2019-10-30 20:23:51

First of all, Elo is for zero-sum games. BB survival is not a zero-sum game you don win or lose to your BB

Yes, the difference between the ratings of the winner and loser determines the total number of points gained or lost after a game.

BUT Your opponent is not the BB, your opponents are the previous mothers of that BB. You don't win against the BB you win against his other mothers.

jasonrohrer · 2019-10-30 20:24:56

Okay, here's a graph of the effects of different K values on the simulation. This is a 100-birth warm-up, followed by tracking score changes over 100 births in 9 performance bands (simulated player who has 10%, 20%, 30%, ..., 90% of their babies die in infancy, and the rest live until 60).

For a 10% player, the average should be 54. For a 90% player, the average should be 6. So for each K, we can compute the actual average score over 100 births, and see how far we are away from the expected average. Too much deviation from the expected average means than in 100 births, this K value is causing convergence to be too slow.

Then we can also look at overlap between performance bands , which is bad. We don't want the 10% player to have much score overlap with the 20% player. We can compute the total overlap between all bands bands (how far is the 10% player's lowest score below the 20% player's highest score).

We want to minimize both measures. We wand to be as close to the expected average a possible after 100 births, and we want as little overlap between bands as possible.

Here's the graph:

jcwilk · 2019-10-30 20:26:15

Kinrany wrote:

As I understand it, genetic fitness is a way to incentivize players to do the correct thing. There's no point in comparing players to each other, right?

I mean, that's easier said than done right? How is "the correct thing" defined? Why would we not want to incentivize people to do better things than anyone else is currently doing, rather than just "a correct thing"? If it's used for something like eve selection then the question it's answering is basically "which player among the available players is most capable of doing the correct thing" rather than "which players among the available players generally behave correctly". From Jason's reply above it sounds like he's wanting to chiefly differentiate players:

jasonrohrer wrote:

So here's the thing, though... we say, "Do we want the good player to always have a higher ranking than the bad player?" Yes, it seems like we do.

That being said, yeah the elo chess thing is a bit of a stretch since this isn't direct competition, but it's still competition because players are vying to prove themselves to be among the most capable and worthy for whatever reward the genetic fitness ends up awarding... So I think it still fits, but again I'm a little out of my element with hard statistics and game theory.

jcwilk · 2019-10-30 20:28:54

miskas wrote:

BUT Your opponent is not the BB, your opponents are the previous mothers of that BB. You don't win against the BB you win against his other mothers.

This seems interesting... I wonder how that would work, technically it could mean that if you're a baby's first mother and you do a shit job and then the next 10 mothers do a great job with that baby then your rank could subsequently drop even long after the relevant life. I'm guessing that would break the transparency issue though

Thaulos · 2019-10-30 20:29:27

Could you start everyone on 30? Then only half of the way to go.

Last edited by Thaulos (2019-10-30 20:29:36)

jasonrohrer · 2019-10-30 20:30:53

Kinrany and Miskas, Elo isn't "for" anything. It is a mathematical construct with certain properties. One of those properties is that it reacts more to surprising results than expected results.

In fact, it is defined in terms of "score" and says nothing about zero sum. How far is your "score" in this game different from your "expected score." That's Elo.

It is actually a bit of a hack to use it for Chess, because your score is always 1, 0, or -1. Your expected score is some fractional score, based on the strength of your opponent.

In the Elo-like construct I'm using, the "genetic score" converges over time to your average offspring lifespan, which means that your "genetic score" and your "expected score" are one and the same thing.

jasonrohrer · 2019-10-30 20:32:50

Yes, Thaulos, that is possible to assume everyone is a 30 to start.

However, it would also mean that beginning players would go down right away, which isn't so great.

miskas · 2019-10-30 20:40:33

The current Gene Score suffers from one thing. The tragedy of commons.

In reality, our efforts don't really matter cause the only way to reach 60 and stay at 60 is the whole community to live to 60 ... High ranking Chess players don't go from masters to average and masters again. they reach the top and stay there till they age.

For any of us that reached the top 5%, it wasn't because of him but because he happened to be at the margin of error that enough pro players were born to him.

think of the last /die player in a game that all Live to 60 and none /dies.
What would be his score closer to? Whatever he does all his BB will give him the score he needs.

Last edited by miskas (2019-10-30 23:10:55)

jasonrohrer · 2019-10-30 20:41:58

Here's the same experiment performed 1000 times and averaged, to remove the noise:

jasonrohrer · 2019-10-30 20:58:33

Here's a close-up of the minimum point, averaged over 10,000 runs of the experiment:

Here we see that the value that minimizes the sum of these two "bad" quantities is K = 46.5

Such a value gives us 18.100910 total deviation from our expected averages over 9 performance bins (2 years per bin) and a total overlap of 24.233646 between all 9 bins, or 2.7 points per pair of bins.

Running the original experiment with K= 46.5, we get:

Baby death rate 0.1, Lowest score seen = 46.299969, highest = 53.600422
                     Overall Ave = 50.633406
                     Desired Ave = 54.000000
                     Ave error = 3.366594

Baby death rate 0.2, Lowest score seen = 38.030238, highest = 46.713251
                     Overall Ave = 42.955051
                     Desired Ave = 48.000000
                     Ave error = 5.044949

Baby death rate 0.3, Lowest score seen = 36.841250, highest = 44.642360
                     Overall Ave = 41.380202
                     Desired Ave = 42.000000
                     Ave error = 0.619798

Baby death rate 0.4, Lowest score seen = 32.483882, highest = 37.837159
                     Overall Ave = 35.150753
                     Desired Ave = 36.000000
                     Ave error = 0.849247

Baby death rate 0.5, Lowest score seen = 22.793872, highest = 37.512631
                     Overall Ave = 29.717380
                     Desired Ave = 30.000000
                     Ave error = 0.282620

Baby death rate 0.6, Lowest score seen = 17.212986, highest = 23.922422
                     Overall Ave = 21.178241
                     Desired Ave = 24.000000
                     Ave error = 2.821759

Baby death rate 0.7, Lowest score seen = 15.219860, highest = 23.248218
                     Overall Ave = 18.999251
                     Desired Ave = 18.000000
                     Ave error = 0.999251

Baby death rate 0.8, Lowest score seen = 6.897707, highest = 15.069428
                     Overall Ave = 10.491057
                     Desired Ave = 12.000000
                     Ave error = 1.508943

Baby death rate 0.9, Lowest score seen = 3.802449, highest = 8.764275
                     Overall Ave = 5.551406
                     Desired Ave = 6.000000
                     Ave error = 0.448594

K=46.500000, Total ave error = 15.941755, Total overlap = 23.206138

So you can see that during 100 births, (after a 100-birth warm-up), we had almost no score overlap between the 0.1 and 0.2 tiers. This is an excellent result. The players are well-differentiated.

Saolin · 2019-10-30 21:01:21

jcwilk wrote:

This seems interesting... I wonder how that would work, technically it could mean that if you're a baby's first mother and you do a shit job and then the next 10 mothers do a great job with that baby then your rank could subsequently drop even long after the relevant life. I'm guessing that would break the transparency issue though

Usually elo style rating systems don't retroactively adjust scores. It just uses past results to generate the score, not future. So if you were the first mom a new player has they would have whatever the default starting score is and future results wouldn't impact how that initial interaction was scored. If for some reason initial score was way off (say experienced player with a new account), this would skew the results short term, but long term both scores would converge to their "true" value anyway. Think if I got boosted up to a 48, but it "should" be a 45, I will lose more points and gain less on subsequent trials than if it had been the "correct" 45 initially.

Dodge · 2019-10-30 21:02:23

Maybe it should calculate average lifespan of birthed player and give points according to that.

For example a mom gives birth to a player that has an average lifespan of 20 years, if she's able to make him go past 20 years then she earns points if he dies before 20 she loses points.

But with how the game is currently 4 year olds leave and you dont have much control over wether they live or die anyway so it's still random in a sense...

Unless you lock them up in a prison and farm them for genetic score

One Hour One Life Forums

#26 2019-10-30 19:27:22

Re: If genetic fitness actually matters, is life limit still necessary?

#27 2019-10-30 19:30:24

Re: If genetic fitness actually matters, is life limit still necessary?

#28 2019-10-30 19:32:37

Re: If genetic fitness actually matters, is life limit still necessary?

#29 2019-10-30 19:35:10

Re: If genetic fitness actually matters, is life limit still necessary?

#30 2019-10-30 19:40:42

Re: If genetic fitness actually matters, is life limit still necessary?

#31 2019-10-30 19:45:38

Re: If genetic fitness actually matters, is life limit still necessary?

#32 2019-10-30 19:47:43

Re: If genetic fitness actually matters, is life limit still necessary?

#33 2019-10-30 19:48:09

Re: If genetic fitness actually matters, is life limit still necessary?

#34 2019-10-30 19:54:07

Re: If genetic fitness actually matters, is life limit still necessary?

#35 2019-10-30 20:01:34

Re: If genetic fitness actually matters, is life limit still necessary?

#36 2019-10-30 20:04:25

Re: If genetic fitness actually matters, is life limit still necessary?

#37 2019-10-30 20:05:56

Re: If genetic fitness actually matters, is life limit still necessary?

#38 2019-10-30 20:15:37

Re: If genetic fitness actually matters, is life limit still necessary?

#39 2019-10-30 20:23:51

Re: If genetic fitness actually matters, is life limit still necessary?

#40 2019-10-30 20:24:56

Re: If genetic fitness actually matters, is life limit still necessary?

#41 2019-10-30 20:26:15

Re: If genetic fitness actually matters, is life limit still necessary?

#42 2019-10-30 20:28:54

Re: If genetic fitness actually matters, is life limit still necessary?

#43 2019-10-30 20:29:27

Re: If genetic fitness actually matters, is life limit still necessary?

#44 2019-10-30 20:30:53

Re: If genetic fitness actually matters, is life limit still necessary?

#45 2019-10-30 20:32:50

Re: If genetic fitness actually matters, is life limit still necessary?

#46 2019-10-30 20:40:33

Re: If genetic fitness actually matters, is life limit still necessary?

#47 2019-10-30 20:41:58

Re: If genetic fitness actually matters, is life limit still necessary?

#48 2019-10-30 20:58:33

Re: If genetic fitness actually matters, is life limit still necessary?

#49 2019-10-30 21:01:21

Re: If genetic fitness actually matters, is life limit still necessary?

#50 2019-10-30 21:02:23

Re: If genetic fitness actually matters, is life limit still necessary?

Board footer