The Arcade is the primary IsoForge surface. It has four games, and they are not four skins over one training. Each asks the patient to produce a different shape of force over time, and the evidence on force-tracking tasks is clear that distinct shapes engage distinct motor-control strategies (Amematsro et al. 2025). Picking a game is a clinical decision about which subskill you want to load.
Two things are true of all four. The target is set from the patient's own recent maximum, so the same game is a different physical task for every patient. And the difficulty adapts during play, which changes how you read the results. That second point is the part most worth understanding before you prescribe, so it has its own section below.
The original four modes are still available at any time. They are covered in choosing a classic IsoForge game mode, and the mapping between the two sets is at the end of this article.
The four games
Ember Drift
Force steers an ember down a corridor. Pushing harder lifts it, easing lowers it, and the corridor wanders, so there is no setting where holding still is correct. As difficulty rises the corridor narrows.
What it trains. Continuously corrected submaximal force. The patient has to notice an error and close it while still producing force, which is a different task from settling onto a fixed target and staying there. Because the corridor moves slowly enough to react to, this is largely closed-loop control: the patient watches, corrects, watches again.
What it reports. Tracking error as a percentage of target, computed after the session from the recorded force against the target the patient actually saw. Lower is better.
Best for. Patients whose assessment shows a steadiness deficit (elevated nRMSE or CV) and who have enough capacity to modulate force rather than just hold it. Also a good first Arcade game for a patient who found a static hold tedious, because the task is the same skill with something to do.
Forge Runner
Gaps scroll toward the player and each is cleared with a fast, explosive push. Jump height comes from how hard and how quickly force rises, not from how long it is held.
What it trains. Rapid force production. A patient clears dozens of gaps in a set, which means dozens of small ballistic efforts, a different stimulus from the one or two maximal efforts a formal assessment collects. Rate of force development declines faster than peak strength during disuse and recovers on its own timeline (Ruggiero and Gruber 2024), so it is worth training as its own target rather than assuming it follows strength back.
What it reports. Average impulse per jump in newton-seconds, the force-time integral behind each pulse. Higher is better.
Best for. Patients whose assessment RTD windows lag the uninvolved side while peak force has largely recovered. Also athletes whose sport demands repeated rapid efforts rather than one maximal one. For the assessment-side reading, see what RTD Early actually means.
Spark Pulse
A rhythm game. The patient produces repeated pulses on a beat, and the interval between pulses matters as much as the pulses themselves. Failing to release fully between efforts costs the run.
What it trains. Repeatable pulses and clean relaxation between them. Relaxation is the half of the cycle a clinician usually cannot see and therefore cannot coach. A patient who cannot let go between efforts will produce a rising baseline across a set, and the game surfaces that immediately rather than at the end.
What it reports. Rhythm variation as a percentage, the variability of the interval between pulses. Lower means steadier timing. Relaxation time is reported alongside it.
Best for. Patients who produce a good single pulse but degrade across repeated efforts, and late-phase work where the goal is repeatability rather than magnitude. Explosive sustained and ballistic pulse-like contractions do not yield the same rate of force development values (Kozinc et al. 2022), so a patient who looks fine on a sustained assessment can still be poor at repeated pulses.
Cinder Updraft
An isometric hold protocol in game form. The run works through a ramp, a series of holds at the prescribed force, a staircase of step increases, and gusts that push the target around. Each wave is a set.
What it trains. Sustained graded holds across a full working set, including holding through a disturbance. This is the closest Arcade game to a conventional trapezoid force-tracing protocol, which is the form most of the force-steadiness literature uses (Enoka and Farina 2021).
What it reports. Yank as percent per second, the rate at which force wavered during the steady part of each hold. Lower means a smoother hold. It also reports hold drift, the sag from the start of a hold to the end, which behaves as a within-session fatigue signal and is not something a single peak value will show you. On yank as a smoothness measure and its limits, see steadiness metrics.
Best for. Tendon loading protocols, where the prescription is already a sustained submaximal isometric hold and the practical problem is that unsupervised holds drift off target without anyone noticing. Also early-phase work after a period of unloading, and any patient whose steadiness metrics are the limiting factor on their symmetry index.
Adaptive difficulty, and what it costs
Each game carries one difficulty setting that moves during play. After each rep or segment the game looks at how often the patient succeeded and adjusts, aiming to hold them in a 70 to 85 percent success band. It hardens slowly and eases quickly, on the reasoning that frustration costs a patient more than boredom does.
This is a direct implementation of the challenge point idea: practice is most productive when task difficulty sits near the edge of the learner's current ability rather than at either extreme, and the useful difficulty is relative to the individual, not to the task (Guadagnoli and Lee 2004). A fixed difficulty is right for exactly one patient on exactly one day.
Four practical consequences:
- Difficulty is tracked per patient, per game, and per joint, action, and limb. A patient's right knee setting does not leak onto their left knee or their shoulder. This matters because the whole point of bilateral training is that the two sides are not equivalent.
- It carries between sessions, but each session starts slightly easier than the last one ended. The patient gets a warm-up rather than being dropped straight back into their hardest setting.
- You can turn it off. Fixed easy, normal, and hard settings are available if you want a constant task, for instance when you are deliberately repeating a session to compare it.
- The score stops being a trend. This is the real cost and it is worth being explicit about. If the task gets harder every time the patient improves, a flat score across four sessions may mean no change or it may mean steady improvement against steadily rising difficulty. The reported outcome for each game is computed after the session from the recorded force signal, not from the game score, precisely so there is something comparable to look at. Read the outcome, not the score, when you want a trend. The score is for the patient in the room.
Why the games are not interchangeable
What looks like the same muscle action on the surface can engage substantially different control machinery underneath. High-density recordings from primary motor cortex during continuous force tracking show that distinct force-profile families (static holds, slow ramps, slow sines, fast sines, chirps) occupy distinct neural state-space regions, and that closed-loop reactive strategies suffice for slow-changing profiles while faster profiles require pre-planned, template-based control (Amematsro et al. 2025).
The practical translation is that you cannot substitute a slow tracking task for a fast pulsed one and expect the same adaptation. Ember Drift and Cinder Updraft load reactive correction and sustained grading. Forge Runner and Spark Pulse load pre-planned production, because at the speeds they demand there is no time to react and correct within a single effort.
A useful heuristic: if the patient has time to see the error and fix it during the effort, you are training correction. If they do not, you are training the plan.
Mapping from the classic modes
If you know the original four, this is roughly where they land. The Arcade games are not reskins and the tasks differ in real ways, so treat this as orientation rather than equivalence.
| Classic mode | Nearest Arcade game | What actually changes |
|---|---|---|
| Hold Steady | Cinder Updraft | The target is no longer static. A ramp, a staircase, and gusts are added, and hold drift is reported. |
| Make Waves | Ember Drift | The target wanders rather than following a set sine. Frequency is not a dial you set. |
| Take Flight | Forge Runner | Similar task. The outcome moves from gates cleared to the impulse behind each effort. |
| Rapid Fire | Spark Pulse | The emphasis moves from matching a pulse template to sustaining a rhythm, and relaxation is scored. |
Where you want a clean, fixed, repeatable task with no adaptive layer, the classic mode is the better instrument. Where engagement across a full set is the limiting factor, the Arcade game is.
A starting framework
Three patterns that cover most rehab progressions. None is exclusive, and a single block can rotate through several.
- Steadiness deficit (elevated nRMSE, CV, or yank on assessment). Start with Cinder Updraft to load the hold directly, then move to Ember Drift once the patient can hold cleanly and needs to grade force rather than just sustain it.
- Explosive deficit (RTD lagging on the involved side). Start with Forge Runner. Move to Spark Pulse once single pulses are clean, because that is where repeatability rather than magnitude becomes the limit.
- Repeatability deficit (good single efforts, degrading across a set). Spark Pulse directly. Watch relaxation time rather than the score.
What to do next
- Pick the game from the assessment deficit. Engagement is a tiebreaker between two clinically defensible choices, not the primary criterion.
- Read the reported outcome, not the score, when tracking progress. The score moves with difficulty by design.
- Let the adaptation run for at least two or three sessions before judging it. It starts a new patient in the middle of the range and needs a few segments of evidence to settle.
See setting submaximal targets from a recent MVIC for picking the target force, dosing IsoForge sessions across a rehab block for set count and frequency, and reading IsoForge training summaries for how training data sits alongside assessment data.
References
- Amematsro EA, Trautmann EM, Marshall NJ, et al. Motor cortex flexibly deploys a high-dimensional repertoire of subskills. bioRxiv. 2025. doi:10.1101/2025.09.07.674717
- Enoka RM, Farina D. Force steadiness, from motor units to voluntary actions. Physiology (Bethesda). 2021;36(2):114-130. doi:10.1152/physiol.00027.2020
- Guadagnoli MA, Lee TD. Challenge point: a framework for conceptualizing the effects of various practice conditions in motor learning. Journal of Motor Behavior. 2004;36(2):212-224. doi:10.3200/JMBR.36.2.212-224
- Kozinc Ž, Pleša J, Djurić D, Šarabon N. Comparison of rate of force development between explosive sustained contractions and ballistic pulse-like contractions during isometric ankle and knee extension tasks. Applied Sciences. 2022;12(20):10255. doi:10.3390/app122010255
- Ruggiero L, Gruber M. Neuromuscular mechanisms for the fast decline in rate of force development with muscle disuse, a narrative review. J Physiol. 2024. doi:10.1113/JP285667