Welcome to Smashboards, the world's largest Super Smash Brothers community! Over 250,000 Smash Bros. fans from around the world have come to discuss these great games in over 19 million posts!
You are currently viewing our boards as a visitor. Click here to sign up right now and start on your path in the Smash community!
I never published the data because I never got any real results aside from "they're basically unpredictable." I don't even know if I still have it laying around. I was just curious about your methodology because I found there to only be certain distinct heights, not the 150k that you've found. Like I said though, my methods weren't that rigorous; it was done using screenshots.
Yeah, the tests I did were so disappointing for Marth, I decided not to publish them here because it would help Fox players more than Marth players. Basically if Fox spaces correctly, 1SDI is not enough for Marth to escape the second hit.
What about 2 stick DI or Multiple Smash DI?
1. It's the slightest offset from 0 horizontally that reads inputs. |x| = 0.2875 to about 0.35 with Magus input display.
Thx for the immediate response!
There is no such limitation, else I would have mentioned it in my video on teching.
I didn't figure as much, just making sure that wasn't a contributing factor. I appreciate the response man, and thanks for all you've done for the community!
I think the idea that melee situations can be broken down into a rock-paper-scissor game is shared among all competent players. And from reading a lot of melee-related content, I noticed that many people do in fact understand that you will find an equilibrium distribution on how often to do which option in a specific situation. (though none explicitly named nash-equilibria)About general methodology of theorycrafting neutral game / read-based situations:
I am honestly surprised I haven't seen anyone estimate neutral game situations with Nash equilibria yet, since it is in a way the equivalent to what the flow chart is in the punish game. How to apply it is pretty easy.
1. Write down the options of player 1 in the rows, the options of player 2 in the columns. Note that you want to write down an option that you have to commit to before you can react again, so you don't want to list "shine", but rather list "shine WD back", "shinegrab", "shine nair" and so on as separate options. The options you list should at least take as long as your reaction time.
2. Assign values to the individual option matchups. Please don't use percentage, because it leaves the stage position/followup perspectives out. You could work with "centistocks", in this case taking someone's stock guaranteed with no read required afterwards would equivalent to 100. It takes some intuition to esimate the values, but from my experience having some imprecise values isn't as big of a deal, and you can always adjust them.
Assign the values from player 1's perspective.
3. Use some program or site like this.
In the example, enter the following things:
dimension: [row number] [column number]
payoff matrix for player 1: [your table]
type of game: Zerosum
Submit
Somewhere it should say EP (expected payoff), this is what player 1 gets in average if both sides play correctly.
You should look to the row where it says EE (extreme equilibrium): It gives the probability of using that option in the equilibrium, if there are multiple EEs, everything inbetween is an equilibrium. Write the values for both players down
4. The interpretation: Learn, experiment, adjust
Compare the numbers with the probabilities you would use an option in a certain situation. What should be done differently. You can also analyze different situations with slight adjustments by deleting a few options (for example deleting "roll away" to look at situaions where you are at the edge of the stage). You can also find that the results don't feel right and find the cause for that (false sign on a payoff value, left out an important option, estimated a option matchup wrongly).
I will give a pretty generic example, which is just a proof of concept and should be done differently for actual results:
![]()
I just used +1 for winning neutral and -1 for losing neutral for simplicity reasons. iirc the EP was +0.15
The player in the rows is standing in front of the player in the columns who is shielding.
If we just pretend this would correctly describe the situation, this would be my example interpretation (so don't take it word for word):
Grab is obviously a strong option for player 1, but it isn't used nearly half of the time optimally.
The overusage of grab in this situation seems to be the reason why player 2's opt to use spot dodge here at least sometimes, which would be a rather poor option if player 1 mixes it up better.
Wait in shield is surprisingly (or not?) a good option for player 2, wait in shield here means waiting for one reaction time length so the time you stay in shield in average should not be that high.
It should be noted that the equilibrium is composed of >5 different options, while players often just use 2 or 3.
Roll in could be better if we give it higher values for centerstage if it works, and if roll awawy is no option.
Specific tests to simulate the same on platforms could also be done.
I see your point, and the point of my post wasn't that you should always play according to NE, but knowing it helps in either case. The idea should be to make the individual decisions based on other mindgame cues, but having the NE in mind, having it as backup option if you feel like the other player is reading you, knowing which options not to use, and being able to categorize the opponent's pattern easier (after all, you can also find out what is best against someone that doesn't use one of the specific options). The last one is probably most important since it is often hard to spot what the opponent is not doing when not comparing with something else. The higher the level gets, the more important it gets, and between players that haven't played each other a lot "exploring" (throwing out different, not necessarily optimal options to collect information) is still the way to go.
4 1 9 2 2 5 5 1 5 0
If you read the index in the first post you'd see that uthrow uair works for a large percentage range. Fthrow fsmash and dthrow fsmash work on bad DI. Fthrow dash pivot fsmash works against bad DI too at higher percents.
Thanks to #SplitsOnTrees for pointing out this thread to me. In the peach subforum I wrote a write-up on how game theory can be applied to melee. You can find it here. I've specifically looked at RPS situations after peach hits an opponent on shield with various frame advantages, but the write-up is meant to be a general introduction to game theory and how any melee player can use it.
Kadano's Marth uthrow on Marth chart only covers uthrow uair which often doesn't net you much afterward and is a difficult link. It seems clear to me that if uair combos at a wide percent range, surely full jump fair combos on away DI. Away DI followups from uthrow are easily attainable because they will likely be DIing away to avoid nasty fthrow combos.
1. Playing around in debug mode, it looks like fair combos on DIs away and behind on a similar percent range to uair. However, the link is just as hard as with uair, and it doesn't work on no DI or slight DI.
1. If you miss the link with fair, you end up in a better position than with uair since you can fast fall and autocancel earlier. The DI is easy to react to and you can still go for uair if they don't DI away. Fair is better though and you will catch them DIing away to escape fthrow combos.
Yes.
No, buffered spotdodge puts up shield for 1 frame, which will simply make you slide off in that odd tumble state after sliding off while holding shield.
In the clip you provided, M2K fastfell after wavedashing off (holding down prevents you from grabbing the ledge).
Speculating here, because it depends a lot on percent, weight, fall speed, etc., but couldn't you walk forward and dtilt them right after they slide off?
