
64 players, and what happened at 128
Three clean runs confirmed a 128-client Highlands research workload. Dense fights and higher counts exposed different limits, so the game remains capped at 64 players including bots.
Read post
The decisions, experiments, fixes and creative work shaping SHADOWFRAG. A development journal built from the project’s actual records.
Game and development terms, explained
Three clean runs confirmed a 128-client Highlands research workload. Dense fights and higher counts exposed different limits, so the game remains capped at 64 players including bots.
Read post30 posts

Three clean runs confirmed a 128-client Highlands research workload. Dense fights and higher counts exposed different limits, so the game remains capped at 64 players including bots.
Read post
I changed storage, network waiting and retry ownership to remove work before messages leave. The measured gains belong to specific 64-client tests, with their failed baseline kept visible.
Read post
Highlands bots were fighting in the north and scarcely attacking the south. I traced the imbalance to squad deployment, capture decisions and a bridge their navigation could not cross.
Read post
The sound pass follows actions from preparation to recovery, adds wordless ninja voices and tutorial narration, and introduces a studio for hearing edits in the running game.
Read post
Star-heavy tests showed clients receiving traffic without getting a fresh scene. I traced the delivery stages, kept the changes that held up and retained the mixed follow-up results.
Read post
The game has grown from close arena fights into conquest, training and a story mission. Movement remains the skill that connects those different ways to play.
Read post
Highlands needed useful fights at each settlement as well as routes between them. I built walking connections first, then added exposed skyways for faster crossings.
Read post
A duel needs clear aim and weapon information; Shogun also needs ownership, supply and routes. The HUD editor and tactical map let those needs share the screen.
Read post
Precision Dojo gives you fourteen offline forms for reading the game’s movement, from lateral tracking to airborne fights with your own weapon settings.
Read post
Shogun spreads a team fight across six settlements and twelve outposts. Capturing the supporting ground opens the next attack and keeps reinforcements supplied.
Read post
The Crimson Path introduces movement, weapons and linked captures through a mission. Each new tool arrives with an encounter that gives you a reason to use it.
Read post
Texture Workshop lets me try a surface across maps before applying it. The harder part was making a batch change recoverable without overwriting later map edits.
Read post
Native demos let me change viewpoint and inspect the server’s decisions after a match. They also need to show what was never recorded and what a client report cannot prove.
Read post
The menu had gained more options than its original layout could comfortably explain. I reorganized settings and previews, then checked that precise values survived editing and saving.
Read post
I refined hats, weapons and effects around shapes you can recognize during a fight. More detail only helped when it made the ninja or the action clearer.
Read post
I built the map atlas from engine captures so players could inspect real routes. The original twenty-map work is preserved here, with later updates to the roster.
Read post
An 80-client run completed, but its slower action results made it a poor basis for raising the limit. I kept the game at 64 players including bots.
Read post
The movement worked before the body explained it clearly. I added directional flips and short local effects while keeping animation separate from combat collision.
Read post
I was sending appearance data and empty fields more often than the game needed them. Removing that repetition reduced traffic while preserving the facts used to decide a hit.
Read post
My first model underestimated outgoing traffic in two final runs. I revised it to account for player-to-player traffic and checked it against new configurations.
Read post
Drawing a frame quickly does not tell me how long an action waited to be processed. I separated those jobs, then measured a smaller rendering change on its own.
Read post
A fast fight can hide how the movement works. I built the tour around separate demonstrations, then used a local bot match to show the same actions together.
Read post
A route can look right from above and fail under a player’s feet. Map Studio uses the game’s movement and collision so I can find those problems while building it.
Read post
I wanted to try an object against a route before adding it to the map. Placement now starts with a preview that can be moved, rotated or cancelled.
Read post
The game had grown beyond ninja instagib. Renaming it meant finding room for those new fights while carrying existing settings, packages and tools along with it.
Read post
The bow needed to show its draw without covering the opponent. I revised its first-person framing and checked how much of the aiming area it occupied.
Read post
SHADOWFRAG began with one-shot ninja fights. As I added weapons and objectives, I kept coming back to movement that gives you useful choices in every kind of match.
Read post
A leap, an air step and a shadow warp become more useful when they work together. Here is how I connected the movement controls and the routes built around them.
Read post
Players were covering a lot of ground in a very short jump. I reduced horizontal travel while keeping the height, airtime and movement combinations intact.
Read post
I separated objectives, weapons and protection so you can change the kind of fight on a familiar map. That also meant teaching the bots what each match asks them to do.
Read postNo posts match your search.
Dates follow the development work described. Screenshots have their own archive dates. RSS feed