Top 10 best online slots: five questions are wearing that one word

1

Five questions, one word, and only one answer anybody can spend

Somebody who asks for the best online slots has a question in mind, and it is one of about five. Which of them leaves most of a credit standing once a condition has been worked through. Which costs least for a given quantity of play. Which gives the most to watch per unit of money. Which can produce the biggest single result. Which can be checked most thoroughly before anybody commits to it.

Those are all reasonable, and this card takes each one seriously and works it out on the family of machines it has figures for. The result of doing so is this card’s finding, and it is not a list. Two of the five are the same question dressed differently and produce one answer between them. One cannot be answered at all for any title here, because the quantity it needs is published nowhere. One sorts every title into a clean order and attaches no money to the order whatever. And one gives an identical answer for every title in the family, which makes it useless for ranking and rather useful for something else.

Each sense takes its quantity from somewhere, and naming the source is half the work. The first two take the published return, which the maker prints for every title and the game reports back. The fourth takes the published ceiling on one round, printed in the same place. The fifth does not take a quantity at all but counts them, asking how many of the machine’s properties are published and how many are not. And the third would need a quantity nobody publishes, which is why it alone, of the five, cannot be worked at all. So four of the five senses can at least be computed from documents a reader can open, and the disagreement between them is a disagreement about what to compute rather than about what is true.

2

Best as what survives a condition, which is the sense with money in it

Take the first reading: of two machines carrying the same offer, which leaves more of it standing. This is the sense every page on this domain is built to answer, and it answers in pounds.

A credit of £40.00 with twenty-five times of staking required of it puts £1,000.00 across the reel. Where 96.71 per cent is published the reel’s own share comes to 3.29 points of the whole of it, so £32.90 of that thousand stays where it fell and £7.10 of the credit survives. B here is £40.00 and W is twenty-five, and g stands at one because a slot counts in full: E = B × (1 − W × (1 − RTP) ∕ g).

Read the bottom row before anything else. Nought point six four of a percentage point is the entire difference between the two machines on this axis, and it decided £6.40 of a £40.00 credit, which is sixteen per cent of the face value. The second title’s break-even sits at 25.45 times, so the offer as stated leaves it with less than half a multiple of room; a condition of twenty-six times would put it below nothing while the first title still had four and a half multiples to spare.

One more line of that table deserves pulling out, because it shows how the two quantities divide the work. Where a British licence holds a requirement to ten, the same forty pound credit closes at £26.84 on the first title and £24.28 on the second, so 0.64 points of return is worth £2.56 there. At twenty-five times it is worth £6.40. The title’s contribution has grown by exactly the factor the condition grew by, two and a half times over, because both figures are the same subtraction applied to a different quantity of staking. The published return fixes a rate and the condition fixes how much of it gets charged.

3

Best as what a thousand pounds of play costs, which is the first question again

The second reading sounds different and is not. Forget credits and conditions, and ask instead what a stated amount of play costs on each machine. Say a thousand pounds of staking, reached however a reader likes: ten thousand turns at ten pence, two hundred at five pounds, or any other road to the same total.

A thousand pounds of staking is expected to cost £32.90 where 96.71 per cent is published, and £39.30 where the figure is 96.07. The difference is £6.40.

That £6.40 is the same £6.40 as the table above, and the two £32.90 figures are the same figure as well. It could hardly be otherwise: a condition of twenty-five times on a £40.00 credit is a way of describing a thousand pounds of staking, so asking what the credit survives and asking what the staking costs are one question with the subtraction done in opposite directions. The credit’s survival is the face value less the cost of the play the condition demands.

£1,000.00 of staking costs £32.90 at 96.71 per cent and £39.30 at 96.07, and the £6.40 between them is everything the choice of title is worth on this card

This is worth establishing rather than waving at, because the two senses get quoted as though they were independent tests a machine could pass separately. They are not. A list ordering titles by how well a bonus holds up, and then ordering them again by cost per unit of play, has published one ordering twice. The second one is not corroboration.

4

Best as how much happens, which has no answer for any title here

The third reading is the one most readers actually have in mind, and it is the one this card cannot answer for a single title in the family. The question is how much happens per unit of money: how often something lands, how often the feature round arrives, how wide the swings are between those events.

Not one of those quantities is published. The maker’s pages for the thirty-six titles in this family carry no figure for the width of the swings, and the data the games themselves receive has no field that one could sit in. No base-game hit rate is published for a single title, and no free-spins trigger frequency either, neither on those pages nor in what the games receive. Figures for both circulate, and the ones this publisher found disagree with each other, which is a reason to print none of them rather than a reason to average them.

There is exactly one arrival rate in this family with a primary source behind it, and it is for the largest single round the original can pay: one in 3,880,481 turns. One figure, about one outcome, on one title. What a reader wants from this sense of the word is the common case rather than the rarest one, and the common case is not published anywhere.

So this sense produces no ordering at all. It is not that the family scores badly on it; it is that the axis does not exist in any document anybody can cite. A list ordering slots by how lively they are has either borrowed a figure with no publisher behind it or is reporting an impression, and those two are easy to tell apart because only one can show its source.

5

Best as how big the biggest thing is, which sorts cleanly and values nothing

The fourth reading is the one the advertising uses: which machine can pay the largest single round. Every title in this family publishes that quantity, so unlike the third sense the axis genuinely exists, and it sorts the family from end to end without a tie worth arguing about.

What it will not do is attach money to the order it produces, and this card does not re-open that arithmetic because another page in this section has already done it and found the expected value of ordering by this axis to be nothing. What belongs here instead is the structural reason, which is short.

What is published there is a multiple of whatever the stake is, rather than a sum. The two titles in the table above sit at 2,100 and 10,000 stakes for one round, and neither of those is money until somebody says what a turn costs. At ten pence a turn the second one’s ceiling is £1,000; at five pounds a turn it is £50,000; and the machine has not changed between those two sentences. So the axis answers the question which, and it refuses the question how much, because the missing term is the reader’s own and no title has it.

There is a second thing the axis will not do, and it follows from the first. Because a ceiling is a multiple, it says nothing whatever about the chance of reaching it, and the chance is the quantity that would turn a ceiling into a value. Part four has already established that the family publishes one such chance, for one outcome, on one title. An ordering by ceiling is therefore an ordering by what is possible, which is a perfectly real thing to want, and it is not an ordering by what is worth anything.

6

Best as what can be known in advance, where the whole family scores alike

The fifth reading is the one this publisher respects most and it produces the strangest result. Ask which machine a reader can find out most about before committing anything to it, and the family answers with one voice.

Five quantities are published for every one of them, and they are the same five throughout: a return to player, the shape of the grid, how many paylines pay, what range of stakes the game accepts, and how large a single round can be as a multiple. All five sit on the maker’s page for the title. Four of them come back a second time in the data the game itself is sent, so a reader can set one source against the other instead of trusting either.

And every slot in it is silent on the same two. No figure for the width of the swings. No rate at which anything arrives, bar the one at part four. That silence is uniform: it is not that some titles document themselves better than others, but that all of them stop at the same line.

One wrinkle is worth having, since it is the only thing on this axis that puts any title above any other. Two of the thirty-five slots carry a page figure their own software contradicts, by two hundredths of a point in each case, while the remaining thirty-three match to the second decimal. On the thousand pounds of staking supposed here, two hundredths of a point comes to £0.20. So this sense splits the family two ways and the distance between the groups is twenty pence of expectation, which is the whole ranking it is able to produce.

7

What the five do to each other when a list has to pick one

Set the five side by side and the trade between them is plain. The first and second agree because they are one question, and together they rank by published return. The third ranks nothing, because its quantity is unpublished. The fourth ranks confidently by a quantity every title publishes, and that ranking is worth nothing in expectation. The fifth ranks everybody equal, bar twenty pence between two groups.

Now notice that the first and the fourth can disagree outright, and in this family they do. The title in the table with the higher published return has its ceiling at 2,100 stakes for one round. The one with the lower return has a ceiling of 10,000 times. A list ordered by what survives a condition puts them one way round; a list ordered by what is possible puts them the other way round; and both lists are correct about the question they asked.

Both disagreements can be sized, and sizing them is the only way to tell which one a reader should care about. On the first sense the gap between the two titles is £6.40 of a forty pound credit, which is sixteen per cent of its face value, and it is an expectation in pounds. On the fourth the gap is 2,100 stakes against 10,000, a factor of four and three quarters, and it is a ratio of two published multiples carrying no expectation at all, that question having been settled elsewhere in this section. So the two senses disagree about which title to put first, and they disagree in units that cannot be set against each other. A reader who wants both answers is asking for two lists, and the honest form of a top ten would be to print the question above the names.

There is one small thing a reader can do about that, and it works. Ask which of the five a list has used, and look for the quantity it was computed from. A list built on the first sense will show returns and a condition. A list built on the fourth will show ceilings. A list built on the third will show a figure with no publisher behind it, and that is the one to put down.

8

The honest limit on the only sense that pays

Having argued that one sense of the word is the usable one, this card owes a limit on it, because the figure it produces is routinely read as more than it is.

The £7.10 at the head of this card is an expectation. It is what the arithmetic gives when a published proportion is applied to a stated quantity of staking, and it describes the long run of that staking rather than an evening. A reader who takes this offer once will not finish with £7.10. They will finish with some other number, and the published return has nothing to say about which, because it is a proportion and not a forecast.

What it does say, and this is the part worth keeping, is the direction and the size of the pull. Every pound staked where 96.71 per cent is published is expected to leave 3.29 pence behind it, and that does not shift with how an evening is going. The same arithmetic applied to a bigger multiple gives a worse answer in exact proportion, which is why the condition rather than the title does most of the work on almost every card on this domain.

The proportions on this card make the same point in one line. The condition took £32.90 of the £40.00 and the title decided £6.40 of what was left, so the condition is worth five and an eighth times what the choice of machine is worth. A reader choosing between two titles at a fixed condition is picking over the smaller of the two quantities, and a reader choosing between two conditions on one title is picking over the larger. Every ranked list of slots is an argument about the small number.

9

What a ranked list of ten is actually telling a reader

Suppose a list of ten names is in front of somebody. On the arithmetic above, here is what it can and cannot be carrying.

It can be carrying an ordering by published return, in which case it is sound, it is checkable against the maker’s pages, and the gap between its first and last entries can be priced exactly: for the two titles here, on this card’s supposed condition, £6.40 of a £40.00 credit. It can be carrying an ordering by ceiling, in which case it is also sound, also checkable, and worth nothing in expectation. It can be carrying both at once, in which case the two orderings fight and whichever one the list has put first is doing the work.

It cannot be carrying an ordering by liveliness or by swing width, because no such figure exists for any title here, and a list that implies one has either taken it from a source that does not have it either or has reported a feeling.

And it cannot be carrying an ordering by how knowable the machines are, because on the evidence of this family they are all knowable to the same depth and silent in the same places. A ten built on that axis would be thirty-five titles tied for first, bar two that are twenty pence behind.

10

Five senses, totalled

On what survives a condition: forty pounds asked for twenty-five times over leaves £7.10 where 96.71 per cent is published and £0.70 where the figure is 96.07, which is £6.40 of difference out of 0.64 points. Break-even stands at 30.40 times and 25.45 times, so one title has four and a half multiples of room at this condition and the other has less than half of one.

On what play costs: £1,000.00 of staking costs £32.90 and £39.30 on those same two reels. Identical arithmetic, identical ordering, and the same £6.40, because the two questions are one question.

On how much happens: no answer, for any title in the family, from any source this publisher would print.

On how big the biggest thing is: a clean ordering with no money behind it, because the quantity is a multiple and the stake that would turn it into a sum belongs to the reader.

On what can be known beforehand: five published quantities per title and two silences per title, the same five and the same two for every slot in the family, with two titles twenty pence behind the other thirty-three on their own makers’ disagreement. The one-sentence version, for anybody who came here for a list: the only ordering with a number attached is the one built on the figure the maker prints, and that number is smaller than the one the condition decides.