The number on the box

Play'n GO publish, for each game, every return configuration the game ships in, and they name one of them as the default. Of the 358 titles in their catalogue, 345 carry more than one figure.

For every one of those 345, the named default is the highest figure published. Not most of them. All of them, with no exceptions.

That is worth stating carefully, because it is a fact about their data rather than about ours. The default is read from the studio's own field for it, and a record whose named default is not among the figures it publishes is recorded as naming none, which happened to 2 titles in the whole set. Nothing in our pipeline picks the highest figure and calls it the headline. It came back that way 345 times out of 345.

So the number that leaves the studio, gets into a press release, gets republished by a comparison site and gets quoted back by an AI assistant is, by construction, the most favourable setting the game is sold in. It is not an average of the settings, and it is not the setting most operators run. It is the top one.

If reduced return settings were a per-game design choice, you would expect them to be spread out: a bit of tuning here, a different shape there. They are not.

Round each title's rungs to the nearest whole point and 326 of the 345 multi-setting titles, 94.5% of them, come out identical: 2, 5, 9 and 12 points below the advertised figure. At the raw published precision, 321 of the 342 five-setting titles sit within a quarter of a point of that ladder on all four rungs, and 329 within half a point.

There is a second, independent way to see it. Take the five figures each title publishes and look only at which whole number each falls into. For 326 of the 342 five-setting titles the answer is the same five bands, 96, 94, 91, 87 and 84, in that order. That matches how the figures are stored on the studio's pages, where each setting sits in its own field named for its band, and it is the reason to read this as a product configuration rather than as an artefact of how we grouped the numbers.

The practical reading is blunter than the statistics. There is essentially one ladder, and it is available on essentially every game. Whichever Play'n GO title a player has in front of them, a version of it exists at roughly twelve points less.

The rows below the first are not a second product line. The three shallowest ladders belong to video poker titles, where the maths of the game sets the figures and there are four settings rather than five, and most of the remaining rows hold a single title whose rungs round across a boundary.

A return figure is what a game pays back over the long run. What a player actually pays is the rest of it, and that is the quantity that moves when the setting changes.

At the median, the advertised setting is 96.24% and the lowest is 84.23%. As returns those are 12.00 points apart and look adjacent. As costs they are 4.2 times as much.

The two costs are $3.76 and $15.77 per $100 staked.

Staked, not deposited. A $50 bankroll played at $0.50 a spin for four hundred spins stakes $200, so it carries four times the numbers above in expected cost, and the gap between the settings widens with it. The longer someone plays, the more the setting they are on matters, which is the opposite of the intuition that a couple of percentage points is noise.

The widest ladder among the five-setting titles is Mystery Egg Surprise, published from 96.98% down to 83.99%: $3.02 against $16.01 per $100 staked, 5.3 times the cost.

Everything above is a count over a catalogue, and a count over a catalogue is an easy thing to read past. The only title named so far is the widest ladder in the set, which is the least representative game on the page. So here is the same finding on three titles this studio has shipped for years and licensed widely, picked for being recognisable rather than for anything about their figures.

Rich Wilde and the Book of Dead is published at 96.21% and at 84.18%, and the same game costs 4.2 times as much to play at the bottom of that ladder as at the top. Between those two settings sit three more, and a player meets whichever one the operator running the game chose.

None of the three is an outlier. All of them carry the 2, 5, 9 and 12 point ladder from finding 2, the one 326 of the 345 multi-setting titles share, and their spreads run from 11.89 to 12.03 points against a catalogue median of 12.00. That is the reason to name them. Not that these games are exceptions, but that the games a player is most likely to have opened are on the same price list as everything else in the catalogue.

Which three titles appear here is the one editorial choice in this study. Every other figure on the page is read out of the data; this is a judgement about what a reader would recognize, and it is worth naming as one rather than presenting as a result. What is not a judgement is whether they are representative: the study's own test suite fails if any named title drifts off the modal ladder, so a re-run cannot quietly leave an exception standing here as the example of a rule.

The other studio in the reference, Pragmatic Play, publishes exactly one figure for each of its 639 titles. Their game pages carry a single return line and present it as the game's return.

You might expect that to show up in the numbers somehow. It does not. 87.2% of Play'n GO's advertised figures and 90.8% of Pragmatic Play's single figures land in the same narrow band, 96.00% to 96.70%. The medians are 96.24% and 96.50%.

Put the two catalogues in one table sorted by RTP, which is what a comparison site does, and the ordering is meaningless in a way no reader could detect. One column is a ceiling and one column is a point estimate, and they are printed in the same font with the same percent sign.

What this does not say. It does not say Pragmatic Play ships its games at one setting, and it does not say it ships them at several. We have no evidence either way, because a studio that publishes one figure tells you nothing about whether a second exists. That is the entire asymmetry: one disclosure format answers the question and the other does not raise it.

The lowest figure Pragmatic Play publishes for anything in its catalogue is 91.93%, and that is one title out of 639.

Every one of Play'n GO's 345 lowest settings is below it. All 345. Their floors run from 81.49% to 90.38% with a median of 84.24%, and the whole of that range sits outside anything the other studio discloses for any game at all.

Again, this is a finding about disclosure and not about the games. It does not mean one studio's games return less. It means that if you assembled every published return figure in this dataset into one distribution, you would be looking at a distribution whose entire bottom half comes from a single studio, for the single reason that it is the only one that publishes that part.

Anyone building a dataset of slot returns from studio pages is going to run into this, and the two obvious handling choices are both wrong. Take the headline figure per title and you have a table of ceilings and point estimates mixed together. Take every figure and one studio's reduced settings dominate the lower tail entirely.

The obvious follow-up question is whether the discount is selective: are the better-returning games cut harder, or protected?

Neither. Median spread is 12.00 points among the 329 titles advertised in the 96% band and 11.88 points among the 8 in the 95% band. Every other band in the catalogue holds two titles or fewer and is reported below without being leaned on.

A null result, and it is reported here because it is the one a reader would otherwise assume we had checked and buried. It also strengthens finding 2 rather than sitting beside it: a discount that does not vary with the game is more consistent with a catalogue-wide setting than with anything designed per title.