What Actually Moves Your Capture Rate

Capture power is a plain integer on the sphere item, and it is the same number whether you crafted the sphere or bought it. The three rate columns hold the species weight at 1.0 and vary only the target, so you can read straight across to see whether a tier survives contact with a harder catch.

The practical reading: a sphere that looks fine on a level 30 target at half health is worthless twenty levels later. Between the Hyper Sphere and the Legendary Sphere the same throw goes from 3.37% to 56.22%. Carry the tier that matches the zone you are hunting in, not the tier you can afford to mass produce.

Every effigy rank adds a flat 0.5 capture power, and capture power is a single summed number: the model cannot tell where it came from. So a Pal Sphere at effigy rank 14 is not approximately a Mega Sphere, it is mathematically identical to one. These pairings are exact identities, not fitted comparisons, and they hold at every level and every HP value.

Target level enters the chain through a steep decaying curve, and it is the single largest term in the model. On a Pal Sphere at 50% HP, the same throw that lands 41.19% of the time on a level 1 target collapses to <0.01% by level 50.

Nothing on this page rescues a throw that is three tiers under-gunned. Sphere tier has to keep pace with the level of the zone you are in, and when it does not, no amount of chip damage or effigy grinding makes the difference back up.

At the reference target, taking a pal from full health to a sliver moves the throw from 0.17% to 63.32%. It costs ammunition and patience and nothing else, it works on every species, and unlike the sphere it does not run out.

Note the shape: most of the gain is already banked by the time you reach 25%, and the last stretch from 10% down to 1% buys only a few points. Grinding a boss to a literal sliver is a good way to kill it by accident for very little extra chance. Stop somewhere around a quarter health and start throwing.

Almost every input in the chain is a multiplier, which means each one is worthless once another has already driven the product to nearly zero. The back or leg hold is different: the native constant adds a flat 0.3 to the rate. Additive terms survive a collapsed chain, so this is the only lever here that still returns a real number on a target you have no business fighting.

These are worth understanding precisely so that you stop trying to play around them. Same reference target, same method.

1. 1Target levelModeled+99.3 pp
   Level 500.16%→Level 199.50%

   The single largest term in the whole model, and the one you cannot negotiate with. A wild spawn of the same species at a lower level is a different problem entirely.

2. 2Critical captureNative value+47.7 pp
   No crit22.56%→Crit procs70.29%

   A flat x2 on the rate before the display transform. It rolls at random, the aim display never includes it, and you cannot force it.

3. 3Species capture weightNative value+36.9 pp
   0.8 (Mimog)12.93%→1.5 (Lamball, Cattiva, Chikipi)49.80%

   Every pal carries its own multiplier. 26 of the 288 pals sit off the neutral 1.0 value, so the spread is real but narrow.

4. 4Alpha / field bossModeled-13.5 pp
   Ordinary pal22.56%→Alpha variant9.02%

   Alphas take a penalty multiplier on the whole chain, which is why a boss arena fight ends in a fistful of wasted spheres if you brought the wrong tier.

The species weight surprises people most. It is narrow in spread and wider in effect than that suggests: only 26 of the 288 pals carry anything other than the neutral 1.0, and the whole range runs from 0.8 to 1.5. Priced against the lever above it: clawing the worst weight back to a neutral pal at the reference target takes 5 effigy ranks. A stubborn species is a real cost, just not one you can do anything about at the moment of the throw.

The hardest realistic case: a level 50 Alpha, beaten to 1% health, wearing the boss penalty. Bare sphere versus a fully kitted throw (effigy rank 15 plus a Capture Strength +3 module) versus that same throw taken from behind.

Palworld’s capture routine lives in native code and has never been published. What we have is a set of exact constants recovered from the game plus a curve fit to a datamined oracle, and mixing those two without saying so would be the easiest way to mislead you. Everything in the table below is verbatim.