Deliberate Practice in Yomi

[vengefulpickle]

I was watching some Prismata analysis recently, and the topic of deliberate practice came up (in this video, if you’re curious). I’m curious to apply it to Yomi (and in particular explore ways to practice effectively outside the tournament setting).

The core pieces of deliberate practice are (from http://expertenough.com/1423/deliberate-practice):

1. You must be motivated to attend to the task and exert effort to improve your performance.
2. The design of the task should take into account your pre-existing knowledge so that the task can be correctly understood after a brief period of instruction.
3. You should receive immediate informative feedback and knowledge of results of your performance.
4. You should repeatedly perform the same or similar tasks.

@cpat was helpful enough to provide some of what he sees as the key things that have helped him in Yomi on discord the other day, as well:

1. I think my grasp on valuation is really strong. I know what the “right” option is pretty much all the time, for both myself and my opponent.
2. Despite this, I have a good handle on when the appropriate time is to steer away from the “right” option, which increases my unpredictability. (e.g. hey I’ll play a random 9 here)
3. I think I am really good at hand-reading, that is, determining what the best options in someone’s hand are and what their short-term combat goals are.
4. I’m very good at switching ranges on the fly.
5. Zane is a very good character that enables 1-4 really well.
6. I’ve played this game for over seven years. I’ve been in every situation before so I don’t play with any emotions. I have no fear of losing a game, because I’ve already lost many times. This makes my plays less exploitable than someone who always plays the safe option, or the risky option. Additionally, because I’m very experienced in the game, it’s easier for me to pick up on certain tendencies.
7. I can quickly calculate the odds of a specific card or combat option in hand because I check the opponent discard pretty much every turn, and use this to guide my decisions.

I’d love to figure out some exercises based on this list that would be playable with the Yomi AI. Thus far, I’ve thought of

1. Identify the Mixup: Pick a high level player, and load one of their replays. Observe their opponent. While stepping through the replay, before each combat reveal, record the top 3 actions you think your opponent might play. These should be as specific as possible while remaining functionally equivalent. So, you might list “Dodge into Anarchy/Raw Anarchy/Counterthrow [5/7]” against a late-game Zane player. After the combat reveal, record what theopponent played. Grade yourself on the number of plays where you correctly predicted the opponents range (they played one of the options you listed). This covers skills 1 and 3 on @cpat’s list, I think.
2. Range Shifts: Pick a replay as in Identify the Mixup. After each combat, decide if there has been a range-shift, and if so, record what the new range is. Also record each combat reveal by the opponent. After the game, grade yourself by counting the numbers of blocks/throws/attacks/dodges in between each range shift you identified. Did they match your expectations? Why were the ranges your predicted incorrect (if they were)? What was the opponent trying to do instead?
3. Play Along: Find an interesting hands-off video record (for example, from here, here, or here, for example). Mute the commentary. As you watch, record your guesses about what each player will combat reveal, and what each player revealed. Pause the video if the players are playing to quickly. At the end, evaluate how you did on each turn. Why did the players play what they did? Would your plays have been better (both locally, and globally)? Source: @mysticjuicer
4. Commentary Analysis: Choose a recording as in Play Along. Leave the commentary on. Listen for the commentators certainty about what they know, or what they suspect. Did you know/suspect those things as well? If not, go back and figure out how they knew (or why they suspect what they did)? Are there types of inference that you are consistently missing? Source: @mysticjuicer
5. Checklist: Write out a list of actions you want to take each turn. Play through some no-timer games and record the result of each checklist item each turn. At the end of the game, check: did you miss any? Why? What was happening? For an extra challenge, explicitly use the checklist during a timed game to force yourself to work through it quickly. An example checklist:
   1. Check both discards, be sure to actively count cards of interest
   2. Consider the exact contents of your opponents hand
   3. Actively label what you think your opponents range currently is
   4. Actively label what your own range is
   5. Identify which options in-hand are likely to win? Are they resources you want to spend right now, or save for later?
      Source: @Shax

Anyone have any other suggestions?

[CKR]

I think the best thing you can do is develop an understanding of the game flow. For me, this came with experience in real matches. I play the AI all the time. It is too predictable, though.

For instance, if the cpu Zane has a super, you better believe it is going to use it. In tournament play this happens much less often. Zane is more likely to throw you. The AI is good for developing an understanding of the game’s systems as long as you understand it’s limitations.

The best thing that I know to do is to look up your opponent’s profile. Look at how they handle situations during each phase of the game. If you can pick up any tendencies, then you will most likely be able to exploit them for the win. This is easier said then done, but it does help.

[vengefulpickle]

Yeah, I kinda figured the bot wasn’t good enough to be really effective. But then I realized that you could easily do the same exercise with the replay of a high-level game, which would probably be much more effective (and opens up a few other avenues for exercises as well).

[mysticjuicer]

It’s important to practice vs the bot just so you can build the skills necessary to win against other beginner players. It also helps break the beginner mentality of “playing to win individual combats” or “predicting the actions of my opponent.” Those are important things to consider, but they will build bad habits in the absence of a good foundation in valuation.

[Hobusu]

So what you’re saying is that everyone needs to grind out unlocking all of the EX characters as part of improving at Yomi?

[vengefulpickle]

The Grave Yomibot just countered my Crashbomb after blocking it…

[Hobusu]

Wow, that’s impressively dumb. Yomi equivalent of a taunt, maybe?

[vengefulpickle]

Well, it did manage to dodge my Max Anarchy later in the game, so it wasn’t entirely on the fritz.

[Shax]

Related to your 1. Identify the mixup: watch your own replays. It’s a tradition in competitive games and sports since basically forever. Recording the exact moves of a board game to be able to replicate and study a match is an early example. Watching high level players is fine and dandy, but it’s just as important to directly address your own play and style as well. I honestly think @mysticjuicer’s success is linked to his youtube work because it’s a perfect combo of analytically watching and reviewing game play of himself and others. #3 on that list seems to indicate that you should always attempt to go beyond theoretical analysis of your own(or a top players) plays to cement the knowledge.

Yomi AI is pretty bad for practice imo. You’re probably better off trying to find a practice partner.

One thing I’ve had the mind to do, and might actually try now, is a checklist method. Write down, in clear steps, things you feel you should be doing every turn. Refer to this checklist literally every turn and make sure you’re actually doing it. My personal checklist would have me specifically considering hand and range valuation questions. Maybe something like:

1. Check both discards, be sure to actively count cards of interest
2. Consider the exact contents of your opponents hand
3. Actively label what you think your opponents range currently is
4. Actively label what your own range is
5. Something involving evaluating my own hand. idk what yet

Any one of these is something that could easily slip through the cracks, and should become second nature over time.

[mysticjuicer]

@Shax

I honestly think @mysticjuicer’s success is linked to his youtube work because it’s a perfect combo of analytically watching and reviewing game play of himself and others.

I would definitely agree. Two similar (but very different) exercises that I recommend is to watch video of games with hands off and “play along.” The faster the players are playing the harder it will be, and of course you don’t get to see the discards unless the person doing the recording checks them, but still. Try to determine what the likely combat reveals are on either or both sides. If you see a combat reveal that confuses you, pause the game and think about what that might tell you about the thinking of each player.

Watching games that are “hands off” is really useful for building your valuation chops, and is the closest you can get to simulating the kind of thinking you are doing in an actual match, where you only have the opponent’s discard and their historical combat reveals to base your decision making on.

Watching games with “hands on” or watching replays in the actual client can also be helpful, because it lets you compare your hand management instincts and combat reveal decision making against those of the player you’re watching.

If you’re just starting off, I would leave the commentary on to try to understand as much as possible what goes into the decisions that the player is making and what they’re basing them on. Keep an ear open for “I know” statements (ex. “I know they have 2 Aces in hand”, “I know they didn’t have any 2s last turn”) and where the player is hedging their bets (ex. “I think this facedown is real, if I play into it X, if I don’t Y, therefore I’ll X.”).

If you’re at the 50/50 winrate level, then I would actually turn the commentary off sometimes, so that you’re purely making decisions based on your own thinking and comparing your outcomes against the outcomes of the player you’re watching, unbiased by the things they’re saying to convince you that their decisions are the right ones.

[vengefulpickle]

I’d like to make the first post in this series a wiki, but apparently I don’t have the requisite permissions. @Leontes, any chance you could help out with that?

[Hobusu]

Just FYI, the first link in Play Along doesn’t do anything.

[Leontes]

This is one thing I love about watching/spectating/casting turn-based games. You get to play along with the person you’re spectating. You get to say “I would Throw here, Leontes will definitely block” with no actual consequences, and then do the same thing next turn, and the next, and you can “see how you did” a bit. Granted the opponent will range shift depending on what other things you would do, but when you are able to identify the mixup that’s occurring, you get multiple reps on guessing that will help you out when you play a real matchup against that character, and maybe even that player!

I also agree that hands-on vs. hands-off are very different experiences. One of my favorite bits of commentary ever is from @MadKing during my IYL2 Grand Finals run.

Listen to what MK says from 23:53 to 26:50:

Watch him call out nearly all of my plays despite not seeing my hand at all.

If the hands were on, it would have been clear to him what my gameplan was in game X; act extremely weak and weave a massive web of valuation traps before entering Dragon Form and going absolutely crazy with my hidden strength. However, the hands were OFF, and he was STILL able to figure this out while commentating the match!

That’s the level you want to be able to reach; with hands-on, you figure out the why of what someone could do, and whether there is any trickery involved in their strategy (I wish I had a good video example of this; maybe I will find one and edit my post later). There are games I’ve played where in very specific circumstances I’ve used 2-Block with when I have other blocks, to get my opponent to shift range more heavily towards throw/reversal on the next turn, or 9-Throw with other Throws in hand to make them feel safe about blocking before they get throwlooped. If you had hands-on while watching me, you’d get to see through my bullshit of whether my hand is strong but I’m acting weak, or my hand is weak but I’m acting strong. A lot of that trickery is how you win with because you can’t always count on your hand being good, but when it’s good, it’s SO GOOD that you have to work extra hard to make it count by getting your opponent to either clam up or overextend at the right time. The satisfaction of getting your stuff to land is what makes me love the character so much.

If you have hands-off, you lose a lot of the magic of what play is really like. You can’t tell if his hand is good or bad. You don’t know if entering Dragon Form with no important stuff in the discard is a ruse, or if it means he’s about to play 4 Kings in a row. This is why I feel can struggle in some situations against even though the matchup is fine; it’s when the contents of his hand are known that let characters checkmate his options, but it’s up to the player to trick them into not knowing where that specific checkmate is.

Lol this turned into a “why I love rant” but honestly this character is the reason I got to where I am in this game. The amount of mental work you have to go through in order to even figure out HOW to win is tremendous, and whatever good stuff I do on any other character is largely due to the Mentor Dragon having taught me a hell of a lot about Yomi.

[Shax]

@Leontes

Watch him call out nearly all of my plays despite not seeing my hand at all.

hmmm, i watched but all he called out was that you were acting weak with a monster hand. He was in fact very surprised by many of your plays and failed to even guess half of them. He even admits it part way through.

On another note I am now considering the merits of blocking with Burst of Speed for the same reasons you blocked with Dragon Form.

[Leontes]

Sure, he wasn’t actually guessing my plays from turn to turn; he was trying to explain both sides of the mixup without taking a side, but his comment about me, specifically as a player, being known to play nutty in order to conceal information was absolutely correct.

IMO his commentary didn’t suggest that he was “surprised” at my plays, (he’s just trying to be entertaining and educational at the same time). The main thing to focus on for me was that he knew why I was playing the way that I was. Sure he backpedals a bit and says “Okay maybe his hand is just bad” but the initial claim that I was full of crap was very very good.

The dJAA combo surprised everyone though, haha.

Anyway, this derail is more in the spirit of “being a good Yomi commentator” vs. “how to deliberately practice” but I think if you can commentate Yomi matches at a high level and accurately (to where other veteran players aren’t really questioning your knowledge) then you’re probably already a strong player.