Penalizing milking on score boards?

[Licorice]

For games where checkpoints can be milked, we can record stock on clear. Then score is score + stock * (highest known checkpoint score gain + k), where k is some constant used to make arithmetic easier.

For example YGW clears R-Type with 1,000,000 points and 10 lives in stock, while ACR clears R-Type with 2,000,000 points and 0 lives in stock. In this hypothetical, we assume the highest known checkpoint milk yield is say 112,889 points (I don't actually know what it is for R-Type, but we can find out by observing a high scoring milking replay), so for the sake of easier arithmetic, we say each life in stock is worth 120,000 points. Now YGW's final score is 2,200,000, and ACR's final score remains the same.

Similarly, for games where bosses (or other patterns) can be milked, we can record time to clear along with score. Then score is score - excess frames over par * score gain per frame on highest yield boss milk.

For example, YGW clears Axelay with 1,000,000 points, and it takes 45 minutes. ACR clears Axelay with 2,000,000 points, and it takes 2 hours. The par time is 40 minutes, and each second over par is penalized by 300 points. Then YGW's final score is 910,000 while ACR's final score is 560,000

Since both the original score as well as the extra information would be recorded on the same score board, both the usual way of interpreting score, and my proposed "fix" could be applied, up to the predilections of the readers.

So the question is, why isn't this ever done?

[Steven]

Because there is probably no point in doing this when you can just have a massive no miss bonus and/or design bosses so that it's not possible to milk them.

[Licorice]

Sure, when making a new game. But what I'm proposing is for existing games where the designers overlooked or didn't care about the problem e.g. R-Type.

[Steven]

It's the same idea in that case. Use a ROM hack to add a massive no-miss bonus, get rid of infinite recovery patterns by reducing the score values they give to be below that of the extend score threshold, and/or redesign bosses so that they can't be milked.

Implementing these changes, or any changes, means you're changing how the game is played, so scores wouldn't be comparable to the original game in most cases (don't think anyone's going to bother separating PCB Kyuukyoku Tiger from M2 Kyuukyoku Tiger, but...), but if all you want to do is get rid of milking, this would work.

[Licorice]

It's the same idea in that case. Use a ROM hack to add a massive no-miss bonus, get rid of infinite recovery patterns by reducing the score values they give to be below that of the extend score threshold, and/or redesign bosses so that they can't be milked.

Implementing these changes, or any changes, means you're changing how the game is played, so scores wouldn't be comparable to the original game in most cases (don't think anyone's going to bother separating PCB Kyuukyoku Tiger from M2 Kyuukyoku Tiger, but...), but if all you want to do is get rid of milking, this would work.

Agreed, which begs the question, why aren't there such "score fix" or "milk fix" hacks?

I guess adding a stock bonus might be more complex (especially displaying it), but conservatively zeroing out certain score values could be an easy win for a number of games.

In any case, in the absence of such hacks, and the absence of the technical skills to produce them, the approach I outlined can be taken by players.

[Kazact]

Agreed, which begs the question, why aren't there such "score fix" or "milk fix" hacks?

Lack of interest, most likely. Some of the newer ports tweak milking, such as Esprade. Anyone can set up their own score track with their own specific rules. I don’t see a reason to mix two rule sets on the same board.

Personally, I’m of the opinion that any milking that has a time out or risk is legitimate score. I may not be interested in doing that particular milking, but the people who are got more out of the game than I did. I wouldn’t want my score normalized based on actions I didn’t actually complete.

[BareKnuckleRoo]

Implementing these changes, or any changes, means you're changing how the game is played, so scores wouldn't be comparable to the original game in most cases (don't think anyone's going to bother separating PCB Kyuukyoku Tiger from M2 Kyuukyoku Tiger, but...), but if all you want to do is get rid of milking, this would work.

I think this is the main reason. Once the game is out in arcades "in the wild" it's not common for it to get major revisions to do things like scoring rebalancing. It'd also be expensive to release revisions for this kind of thing. Milking is one of those things that's accepted as being flawed, but a part of the game, so you deal with it.

You do what you gotta to do get those nice big numbers.

There's also some cases where there's some milking available, but the game infinitely loops and will eventually counterstop anyways, so there's not really any reason to worry about discouraging milking.

[Steven]

Typing on Steam Deck because too lazy to get out of bed to turn on my normal computer, so no super detailed answers right now lol

[
Agreed, which begs the question, why aren't there such "score fix" or "milk fix" hacks?

I can think of some easily: M2's ESP Ra.De., Kyuukyoku Tiger, and most (all?) of their Super Easy modes and arrange modes are just emulated ROM hacks and those alternate modes usually have scoring changes and fixes that you can use in the custom thingy if you want. ESP Ra.De. was the most notable one, as they specifically made Psi to try to fix the original game's broken scoring, although I think it just ended up being slightly less broken.

Their Thunder Force IV and Kyuukyoku Tiger have removed infinite milking in the normal game modes, although they carelessly missed a spot in TFIV somehow. Still not sure where the massive bonus you get from the Toaplan games if you set them to end after completing only one loop comes from, but I think it might be based on how many lives you have when you finish.

Implementing these changes, or any changes, means you're changing how the game is played, so scores wouldn't be comparable to the original game in most cases (don't think anyone's going to bother separating PCB Kyuukyoku Tiger from M2 Kyuukyoku Tiger, but...), but if all you want to do is get rid of milking, this would work.

I think this is the main reason. Once the game is out in arcades "in the wild" it's not common for it to get major revisions to do things like scoring rebalancing. It'd also be expensive to release revisions for this kind of thing. Milking is one of those things that's accepted as being flawed, but a part of the game, so you deal with it.

You do what you gotta to do get those nice big numbers.

There's also some cases where there's some milking available, but the game infinitely loops and will eventually counterstop anyways, so there's not really any reason to worry about discouraging milking.

Almost certainly; there really wasn't that much of a financial incentive just to fix broken scoring back when people actually cared about arcades and went to them, and doing so means you can't compare scores between versions. Now if you make an arranged version with new stuff that could be financially viable, but I doubt it would be worth it just to fix broken scoring. Most would just admit they made a mistake and be sure to not do that mistake again in their next game(s).

[To Far Away Times]

If you can milk in a game it probably isn't worth playing for score.

[Rastan78]

If you can milk in a game it probably isn't worth playing for score.

If you mean infinite milking then I would agree. Otherwise you would be saying the entire Darius series, AC Shinobi, Metal Slugs, Radiant Silvergun, Battle Garegga, Ibara etc etc etc aren't worth playing for score.

IMO it's a personal thing where some will enjoy pushing the limits with milking and some don't. Also i think it can be a bad habit to write off playing games due to observing aspects of high level play from superplays or guides that won't affect 99.9% of players anyway.

Clearing a game consistently with an already maxed out high PB should be reached in most cases before even thinking about longer milking.

A good rule to have is that only an ALL score counts. That keeps the risk/reward in favor of not just tanking all resources on checkpoint milking an opportunity early game, ending credit in stage 3 and calling that community record. (Xexex? lol)

Maybe a rule like only a new #1 score or top 5 or so has to be an ALL. That way to not inhibit newer players from sharing their score progress to leaderboards.

[Meriscan]

I'm very confused by this thread. What problem is this supposed to fix? Why make scoreruns more like survival runs by cutting out a part of the scoring?

If the community thinks certain (infinite) milks make the game unenjoyable to play, then historially we have just said you're not allowed to do it to enter your score on the leaderboard. We do this for Strikers 1945, Daioh and many others I'm sure I haven't seen yet. In scoring competitions like Calice there are clear rules what you can't milk if it would ruin the competition.

People could just say for R-Type or Pulstar that we just add a second category where you're only allowed to milk a checkpoint once or not at all.

I don't think this idea has taken hold on the forum, but you might have seen 'kumites' in the past where usually streamers will do a small competition to see who can clear the game with the most lifes (and bombs as tie-breakers). It sounds to me like you'd rather see something like this than anything score-related, but with an actual leaderboard of course.

[Steven]

I have a very strong feeling that making ROM hacks like this would probably end up being a waste of the creator's time because almost nobody would likely play it anyway unless it got an official release.

[Lemnear]

If you can milk in a game it probably isn't worth playing for score.

But, isn't "milking" a scoring technique?

[Jeneki]

But what I'm proposing is for existing games where the designers overlooked or didn't care about the problem e.g. R-Type.

I don't enjoy checkpoint milking either, but to pull it off a player has to be good enough to no-miss the rest of the game, good enough to recover at the most lucrative checkpoint, and be consistent enough at both to pull it off in the same run. That's a huge commitment and I respect the scores of those to do it even if I'll never attempt it myself.

I'll boss milk Takumi games (Gigawing series and Mars Matrix) because I find the execution to be interesting and enjoyable, and the process doesn't overstay it's welcome. I don't boss milk in EspRade because it's on the longer side and I don't find the execution enjoyable. I'm willing to accept that I'll never be able to compete with the highest scores in EspRade, and I still enjoy playing EspRade despite this.

[Rastan78]

Penalizing milking

There is already a thread called Dick Stock on the forums

[Rastan78]

That's a huge commitment and I respect the scores of those to do it even if I'll never attempt it myself. . .

I'm willing to accept that I'll never be able to compete with the highest scores in EspRade, and I still enjoy playing EspRade despite this.

This is the answer. Every game will have some absurdly difficult, restart inducing aspect whether it's extended milking, rng or something else even more frustrating. Engage with the parts of a game you are enjoying and put the rest aside. Maybe return to it later when you need a new challenge. Be proud of your own personal achievements for what they are bc there will always be a better player out there somewhere.

[pja]

Apart from everything else that was already said, are you really surprised that when people compete for score, they use the actual number that the game shows instead of some arbitrary formula? No one is ever going to bother with that. And even if a hack is doing the calculation for you, it still doesn't solve the part that this is some arbitrary change to scoring that everybody would have to agree on (e.g. using the same hack), that's obviously less appealing than competing on the original version.

[Licorice]

Apart from everything else that was already said, are you really surprised that when people compete for score, they use the actual number that the game shows instead of some arbitrary formula? No one is ever going to bother with that. And even if a hack is doing the calculation for you, it still doesn't solve the part that this is some arbitrary change to scoring that everybody would have to agree on (e.g. using the same hack), that's obviously less appealing than competing on the original version.

Yes, I am surprised because I assume the idea is to measure player skill playing the challenging and interesting parts of the game, not patience with developer oversights.

For example, here's a run of Thunder Dragon 2 achieving 6.204.320 points, exploiting a time out bug on the stage 3 boss (if the first phase times out, the time out for the second phase never triggers) for a 20 minute milk:

Video 1

Here's another run, achieving a slightly lower score of 6.136.880, but without the milking:

Video 2

Which run do you think is more interesting? Which run do you think reflects better play?

[DejahThoris]

Yes, I am surprised because I assume the idea is to measure player skill playing the challenging and interesting parts of the game, not patience with developer oversights.

I don't enjoy checkpoint milking either, but to pull it off a player has to be good enough to no-miss the rest of the game, good enough to recover at the most lucrative checkpoint, and be consistent enough at both to pull it off in the same run. That's a huge commitment.

This is so well put, and responds 100% directly to your statement. Being able to pull off the milking in a lot of these titles IS a measure of skill. It's much harder to do the hard part from a checkpoint with no upgrades 5x in a run than it is to do it once with upgrades.

[Lethe]

For example, here's a run of Thunder Dragon 2 achieving 6.204.320 points, exploiting a time out bug on the stage 3 boss (if the first phase times out, the time out for the second phase never triggers) for a 20 minute milk:

Video 1

Here's another run, achieving a slightly lower score of 6.136.880, but without the milking:

Video 2

Which run do you think is more interesting? Which run do you think reflects better play?

Very dumb example. The runs are with radically different ships, of course they're going to look different. As an observer who doesn't understand every nuance of TD2 you have no way of telling which represents the game better.
The WR with 1P is 6,447,380 and with 2P 6,247,270. Just off those numbers, I could conjecture that the 2P 6.204 run is the more impressive irrespective of any techniques it uses. I could also conjecture that 2P is harder to play than 1P at a high level, because it's much weaker than 1P and involves maneuvering TD2's fatass hitbox around at high speeds, and it might milk worse too. But I really have no idea what I'm talking about, and you don't either!

[Rastan78]

This is kind of shifting goal posts because OP didn't mention bugs and exploits It was just checkpoint milking and bosses than can be milked which is a very broad category.

For something like this wouldn't it be easier to have a scoring category banning the bug than using a convoluted score calculation system?

[Lethe]

In the specific example above, it still doesn't matter because TD2 isn't banned from JP scorekeeping, which means the milk can't be infinite.
For the record, basically every game that has truly broken scoreplay is "officially" closed when that's discovered or some arbitrary target score is demonstrated. Communities already come up with their own soft rules for broken games; e.g. Garegga or the recent Gun.Smoke Calice play. People understand this already, OP just doesn't like milking and is pretending that means something.

[pja]

Which run do you think is more interesting? Which run do you think reflects better play?

Frankly, I don't think it matters because that's not the point of leaderboards Sure you could try to come up with some esoteric formula that best captures the "quality" of a given run but that can quickly get subjective (e.g. what's the exact multiplier we should use? what if the game has no bomb bonus or remaining lives bonus at the end, should we add that in? surely runs where someone died less or bombed less is better?) But at this point it's not really competing in the original game anymore.

And if we're being honest, for any given game there already aren't that many people seriously competing for score - and if they do, chances are they already like the game as it is, despite shortcomings. But still, the main reason I would guess is that it's too complex and arbitrary (and I dunno, it would kinda suck that after the run instead of popping off after that sweet score you just got, you have to whip out the calculator, but I guess that's just part of what I mean by "too complex").

[Licorice]

Why make scoreruns more like survival runs by cutting out a part of the scoring?

This is so well put, and responds 100% directly to your statement. Being able to pull off the milking in a lot of these titles IS a measure of skill. It's much harder to do the hard part from a checkpoint with no upgrades 5x in a run than it is to do it once with upgrades.

You both bring up the very valid point that in many games milking is a feat, risk, and enjoyable challenge, and that the skill ceiling is lowered if you remove these kinds of milking opportunities. The REAL FIGHT AREA 2 milk in the second loop of Image Fight is a particularly strong display of skill and strategy, and adds to the game vis à vis the alternative of a stock bonus given on clear. But IMO it's rather the exception that proves the rule. The kinds of milking I'm talking about are like the example from Thunder Dragon 2 I posted earlier, or, for another example, the notorious blue plane milk in stage 2 of Gun Frontier.

If the community thinks certain (infinite) milks make the game unenjoyable to play, then historially we have just said you're not allowed to do it to enter your score on the leaderboard. We do this for Strikers 1945, Daioh and many others I'm sure I haven't seen yet. In scoring competitions like Calice there are clear rules what you can't milk if it would ruin the competition.

People could just say for R-Type or Pulstar that we just add a second category where you're only allowed to milk a checkpoint once or not at all.

For something like this wouldn't it be easier to have a scoring category banning the bug than using a convoluted score calculation system?

The disadvantage of this approach is that it's difficult to tell if something was a legitimate miss or a milk disguised as a miss. You could set the rule that a miss at a particular checkpoint is always a milk and thus invalid, or allow at most one miss, but that would invalidate some otherwise good runs e.g. a player who just happened to get 2 legitimate misses at a well known milking spot, but otherwise played beautifully everywhere else. Better, IMO, is just to make a life in stock worth a smidgen more than the highest known milk yield.

People understand this already, OP just doesn't like milking and is pretending that means something.

I'm just suggesting a generic way of approaching the problem that has some advantages over e.g. other ad-hoc community solutions such as the ones described above by Meriscan.

Very dumb example. The runs are with radically different ships, of course they're going to look different.

You are right, my comparison wasn't very good. However, as a hypothetical, do you think that a player achieving a score just under the milking score with the same ship, but without the milking, is a worse player than the player who milked for 20 minutes, then got a much lower score on the other 7 stages of the game, but still got the higher score overall due to his milking patience?

[Rastan78]

do you think that a player achieving a score just under the milking score with the same ship, but without the milking, is a worse player than the player who milked for 20 minutes, then got a much lower score on the other 7 stages of the game, but still got the higher score overall due to his milking patience?

I would let go of the idea that scores of isolated runs will ever be a fully reliable measure of who is the "better" player particularly at mid level scores. Shmups are full of low skill high scoring techniques right alongside risky, dangerous, low point yielding strategies.

The problem with a lesser skill player who prioritizes the low hanging fruit and goes for the easy milk is not their score is artificially high. It's that they will likely burn out and hit a plateau much more easily than the player who takes a more balanced approach and introduces milking only to reach a fully optimized PB.

Once you break out the calculator to arbitrate who is the true and honorable winner what next? Now you have to factor out RNG. By the same logic looking at two isolated runs player A has better base score and player B has better luck with lower base score and wins.

The problem here again is isolated runs don't tell the whole story. Maybe player A had the luck on many chances and choked due to nerves bc they can't perform under pressure. Maybe player B got the luck bc they had the perseverance to grind many runs and focused more on grinding than optimization and that proved to be more efficient. Normalizing score with a calculator might take away some excitement and strategy for players who actually enjoy a game that way and commit to deal with a certain randomness.

Then we come to tactics that were not foreseen or intended by the development team. Do we take out the calculator again here? How do we know which tactics were unintended? IIRC Esprade milking was known by the team and left in purposely although they didn't think players would actually be crazy enough to exploit it. So how do you draw the line between intended and unintended there?

Maybe the much more elegant solution is try to improve the player and not the game itself.

[Meriscan]

I'm just suggesting a generic way of approaching the problem that has some advantages over e.g. other ad-hoc community solutions such as the ones described above by Meriscan.

What I talked about IS the general solution. Any OP on a scoreboard someone put any effort in will have a 'dont do X infinite milk' if such a thing exists in a game. Your solution of making an exact formula could get endlessly debated on how exactly it should be tweaked. It's completely impractical anyway, given that it would mean everyone would need to have all submission on video so we can check out for long the run was exactly.

But the general point that I admittedly could have put forward better is this first thing I mentioned in my previous post. 'What problem is this supposed to fix?'

Players who play for score will always compare their scores not just to others on the forum leaderboard (especially older shmups tend not to have very competetive submissions), but to scores from people playing around the world. Who would also never use a system like this.

I guess to summerize my point is that a system like you've described has not been used before, not because no one has come up with it, but because it's undesirable.

[Lemnear]

Apart from everything else that was already said, are you really surprised that when people compete for score, they use the actual number that the game shows instead of some arbitrary formula? No one is ever going to bother with that. And even if a hack is doing the calculation for you, it still doesn't solve the part that this is some arbitrary change to scoring that everybody would have to agree on (e.g. using the same hack), that's obviously less appealing than competing on the original version.

If the game is designed a certain way, that's how it should be played, with some rare exceptions.
The score in the end is the only truth.

[Rastan78]

For example, YGW clears Axelay with 1,000,000 points, and it takes 45 minutes. ACR clears Axelay with 2,000,000 points, and it takes 2 hours. The par time is 40 minutes, and each second over par is penalized by 300 points. Then YGW's final score is 910,000 while ACR's final score is 560,000

Since both the original score as well as the extra information would be recorded on the same score board, both the usual way of interpreting score, and my proposed "fix" could be applied, up to the predilections of the readers.

Think about the overall changes this could make to game balance. It's more than just a fix. Change the rule set and you will change optimal playstyle. Imagine routing for something like this where you're playing according to an arbitrary global time limit. You need to get as close to that limit without going over on every run.

Maybe it's now better to speed kill certain bosses to allow time to milk others, negating certain aspects of skiilled scoring in favor of others. What is the optimal time to spend on each boss under a 40 minute limit? How is that tracked by the player? Music cues? Rom hack timer gadget? Wristwatch? Does firing less shots for slowdown management now become a part of fully optimized scoring?

If players get very efficient at speedrunning the rest of a game so as to still allow time for checkpoint milking what then? Scoreboard wipe and now the allowable time frame is reduced to 30 mins? What if a player does zero milking but they prefer a lower power weapon and defensive playstyle which extends boss fights. Is their score also penalized for exceeding time limit? Do players really want to use both a timer and calculator to submit a score that will not be compared with broader community easily? Do we need to go on?

[Licorice]

Think about the overall changes this could make to game balance. It's more than just a fix. Change the rule set and you will change optimal playstyle. Imagine routing for something like this where you're playing according to an arbitrary global time limit. You need to get as close to that limit without going over on every run.

Maybe it's now better to speed kill certain bosses to allow time to milk others, negating certain aspects of skiilled scoring in favor of others. What is the optimal time to spend on each boss under a 40 minute limit? How is that tracked by the player? Music cues? Rom hack timer gadget? Wristwatch? Does firing less shots for slowdown management now become a part of fully optimized scoring?

Yes it would change the optimal play style, but IMO in a good way, certainly improve it from "sit in the safe spot and milk projectiles for 20 minutes". The kind of trade offs you correctly describe under such rules are, IMO, interesting and strategic.

As for tracking, the best would be to use your emulator's frame number display as frame number is slow-down agnostic. This is no good, of course, for people who play on original hardware.

If players get very efficient at speedrunning the rest of a game so as to still allow time for checkpoint milking what then? Scoreboard wipe and now the allowable time frame is reduced to 30 mins? What if a player does zero milking but they prefer a lower power weapon and defensive playstyle which extends boss fights. Is their score also penalized for exceeding time limit?

It's OK if there's allowance for some milking. The point is simply to remove degenerate and, IMO, boring play due to developer oversights, rather than encourage speed running.

What I talked about IS the general solution. Any OP on a scoreboard someone put any effort in will have a 'dont do X infinite milk' if such a thing exists in a game.

Ok, so they can't do an infinite milk, but can they milk for 10 minutes? 5 minutes? This is the problem that my solution addresses.