2025-10-22 Strategy

Overview¶

In this post I explore how to understand how to come up with concrete strategy

To make this tangeible we will need to:

1. Document basic mechanics of the game

2. Pick specific MU and stage to breakdown

3. See how can generalize

For this we will take Lucas on Small Battlefield

Basic Game Mechanics¶

1. hitstun (what is actionable, and when able vs not)?

2. tumble (what is actionalbe, what is fastest action, how long last)?

3. shield and oos options (frame data, eg up smash no shield drop, etc)

4. document LSI vs DI (only horizontal) vs SDI? using away and up for tech situations, etc

5. when squat jump vs double jump frames?

6. frame calculs like in mii sword fighter vod

Understanding frame data:

1. 6 frames of input delay imposed by SSBU engine

2. fastest reaction time = 12 frames, average = 15 frames

3. OOS options have 11frames of shield drop delay, no delay for jump, grab, up smash, up special.

4. any 50/50 or decsion (ie not reaction) takes 10 frames (Hicks law)

Full frame and stats for each MU: link

Methodology¶

This describes how lab & practice (how use and what settings use in training mod)

Practice¶

We will describe how to practice:

1. SDI

2. DI (DI shift for Kaz)

3. Tech Plat

4. Tech Edge

5. Tech Spiking

6. Mashing out of grab

7. KO punch (gazelle, reverse KO punch, etc)

Key Focus:

1. Roll cancel grab and follow ups out of grab -> force opponents to not shield constantly (with our moves on shield, i consistently get punished or player hermits at ledge)

Vod Review¶

Alternis vs Shisui¶

1. Lucas easily hurtbox shifts under side B (do not use side B unless confirm, if burst option will whiff)

Practice Plan¶

1. Trade dash dancing habit for pivot dancing. It practice micro movement and still can frame 1 shield easily (And also looks funny lol)

2. Move back to patient play, when in doubt dont move or touch controls. This mixed with micro movevement should see whiff KO punches go back up

3. dtilt/uptilt and read habit to see how respond

4. dash then walk right out of burst range, similar with pivot cancel ftilt, you react/start this too late

Observations¶

Mac

1. using foxtrot, it takes 4 flicks dash to cross stage, you get to each quadrent, end of stage, middle of plat, mid stage, other mid plat, end of stage.

2. mid plat doesnt have many good follow up attacks (up air, and up smash need to be mid quarters on side of opponent to hit reliablely)

3. learnign turn around up smash, could help with an easy specific spacing of ledge options to then do follow up plat land read

Lucas

1. from ledge, jump can cover entire stage

2. can airdodge as close to mid hieght (not far from plat), and react directional airdodge that will cover distance to other plat

Data Collection¶

ssbu data will aggregate data on, entire cast and their movesets such as:

1. Ultimate FrameData

2. Mac specific data (Eg mac discord data on what can SL)

This can be used to:

1. Easily filter and fastly understand data that could help understand

   1. Safer strategys

   2. Understand why certain interactions are happening (eg is it something need to train to improve on or is physically not possible because of framedata)

2. Programmatically create portions of these docs

Should I document:

1. KO kill confirm & dtilt true window

2. Quick bullet points on how to use data specifically for mac

3. how document https://ultimateframedata.com/stats in consice way and how could be useful for ledge trapping strats

4. tranlsate/add any stats to more visual queues/distance (Eg grab range translate to length of plat)

   1. come up with visual distance (ways to easily tell by stage, using sb battlefield plat as base, and things like what move take to accurately get there the fastest, eg dash covers entire plat, for example to accurately up smash on plat landing when at edge of stage, need get in middle halves which dash will always overshoot the first half)

5. directional airdodge distance (eg lucas airdodge can be over one SB plat and directional airdodge all the way to other plat, so dont want to commit to one plat if catching landing)

6. burst range window

7. areas off stage can cover vs not (should inform safe recovery paths)

8. document dash turnaround (https://ultimateframedata.com/stats) but what does it mean (have glossary for this data so understand what means)

9. Base line interpreation of frames (eg 3 frame move usually indicates combo break, dash/run speed to know if our walk speed faster, so walking is more prefferred strat)

10. per move how can get out (eg jump vs neutral airdodge, etc), know for mac moves (so know what options opponent can chose) and other characters (so understand disadvantage, need to see if that true of move or combo string)

Ideas¶

1. Strats on how to improve catching landings (ways to ensure know when person cant escape certain area of landing - ie you can react, predict landing, etc).

2. Map out if end lag from lucas pk fire is punishable from burst range. This will help break down movement completely (have to know range, frames of dash, run frame by distance, any lag before get out attack out of run, etc, understanding what frame buffer need to add to make practically, as cant expect to be frame perfect, eg when pk fire comes out when are we usually reacting, accuracy of stopping and getting hit out

3. have some parts of docs auto generated from spreadshet and/or functions/variables from mac python library

Questions¶

1. is turn around up Smash viable? if does that make from ledge to cover near SB plat possible to walk up Smash or dash turn around up smash to cover all options on reaction?

Notes¶

1. this link at 1:57:46 has good example of SL ledge trap (using full hop side B to catch jump, then releasing for neutral get up)