"Scientific American: Why Probability Probably Doesn’t Exist" Topic

link

Life is uncertain. None of us know what is going to happen. We know little of what has happened in the past, or is happening now outside our immediate experience. Uncertainty has been called the ‘conscious awareness of ignorance' — be it of the weather tomorrow, the next Premier League champions, the climate in 2100 or the identity of our ancient ancestors.

In daily life, we generally express uncertainty in words, saying an event "could", "might" or "is likely to" happen (or have happened). But uncertain words can be treacherous. When, in 1961, the newly elected US president John F. Kennedy was informed about a CIA-sponsored plan to invade communist Cuba, he commissioned an appraisal from his military top brass. They concluded that the mission had a 30% chance of success — that is, a 70% chance of failure. In the report that reached the president, this was rendered as "a fair chance". The Bay of Pigs invasion went ahead, and was a fiasco. There are now established scales for converting words of uncertainty into rough numbers. Anyone in the UK intelligence community using the term ‘likely', for example, should mean a chance of between 55% and 75% (see go.nature.com/3vhu5zc).

Attempts to put numbers on chance and uncertainty take us into the mathematical realm of probability, which today is used confidently in any number of fields. Open any science journal, for example, and you'll find papers liberally sprinkled with P values, confidence intervals and possibly Bayesian posterior distributions, all of which are dependent on probability.

And yet, any numerical probability, I will argue — whether in a scientific paper, as part of weather forecasts, predicting the outcome of a sports competition or quantifying a health risk — is not an objective property of the world, but a construction based on personal or collective judgements and (often doubtful) assumptions. Furthermore, in most circumstances, it is not even estimating some underlying ‘true' quantity. Probability, indeed, can only rarely be said to ‘exist' at all.

I thought this might be a thought-provoking article.

Almost every game system uses some type of probability system, such as dice, cards, etc., to determine the chance of an action occurring within a game turn, which can be defined as a specific amount of time or is totally abstracted.

Unfortunately, I've seen many occasions when the odds don't play out as predicted. For example, a unit activates on a 1-4 but four times in a row the player rolls a 5 or 6. Has that ever happened to you? Is there a historical or real-world reason why that happened or do I have to use my imagination? This can be very frustrating for a player and contribute to the game being "unrealistic" or a poor design.

It's been said that all combat is "Time Competitive" which is why the military trains and drills the same thing over and over again so it becomes second nature.

Training standards are determined by how quickly a unit can execute its order. They use Immediate Action or Battle Drills so troops can immediately execute an order or action under certain conditions without needing an order. If you are ambushed, you don't need someone to tell you to hit the deck and return fire.

By knowing the historically average time to execute an order based on training level and weapon platform performance you can establish a baseline for performance in seconds, minutes, hours, or days depending on the scale of the game. Real life is not entirely random. However, how long something will take can be somewhat or very random.

Like my wife telling me to pick up my dirty underwear off the bedroom floor. She might as well roll a D6 to determine how many days it will take.

Certain conditions such as weather, poor supply, untrained crews, or poor leadership/intel can extend the amount of time. Good leadership, veteran crews, or good supply and other factors can increase efficiency and decrease the amount of time.

A die roll can modify the result plus or minus to make it more variable and account for the Fog of War. You don't want it to be too predictable. Enemy action, feints, false intel, suppression, and interdiction can also increase the amount of time.

The above is the approach I take. I also allow players to use historical Risk-Reward Tactical Decisions that allow them to execute more quickly but with a potentially worse outcome. They can also spend more time for a better outcome but risk the enemy seizing the initiative and executing/shooting first. By making tradeoffs, the player cannot put all of the blame on the dice.

Wolfhag

What are probabilities based on? Professional card players understand probabilities based on facts such as a known number of cards in a deck, and a known number of each face value and suit. What is a 4 in 6 probability of movement based on? We see multiple instances in the ACW (probably all wars) in which units do not step off when they were supposed to, frequently failing to meet their time table by many hours. If the 79th Pa is supposed to march at 6:00 AM but doesn't start moving until 3:00 PM, how do we account for that? Is that ten 5 or 6s in a row.

Probabilities are also based on data points. The more you have, the more accurate your estimate of probability. Rolling a d6 ten times will not get you a true probability outcome. Rolling the same die 1,000 times will get you much closer to seeing the true probability of rolling a certain number.

The original quote correctly recognises that language interferes to a significant degree in the misunderstanding/misinterpretation of probability. Then goes on to show that they haven't actually understood what they just said by quoting examples that are not governed by mathematical probability rules.

PA is right in what he says about more data giving closer estimates but the roll of a die is (not unreasonably but not entirely accurately) fixed for each roll but is not a fixed sequence for a series of rolls. 1,2,3 is as likely to happen as 6,6,6.

Regardless of whether the probability can be meaningfully decided upon, which is harder the number of variables that go in to determining that outcome. The way probabilities are presented in most games don't
A reflect how most humans consider probabilities
B encourages a way of thinking about probabilities, that doesn't align with how the actual people think of them and leads to other issues.

And considering that Kennedy pulled the air support as well as other support from the BoP forces, it's no wonder it was considered a fiasco.

Oh, goodie.
Since dice are no longer relativity speaking relevant, I can just declare that I won and go home.

Non-paywalled version:

Archive.Today – Why probability probably doesn't exist (but it is useful to act like it does)

---

PS I hate the concept of "activation" – smacks of someone back at brigade with a bubble-gum chart of the battle.

Possibly better:

- Ready-to-go test – with failures such as "waiting on fresh battery/ammunition/fuel", "waiting on Simpkins to tie his boot laces", etc.

- Did-they-go test – with failures such as "sorting out conflicting orders", "went off in wrong direction", "M60 fell to pieces", "Simpkins tripped up", etc

---

I like the idea of increasing the possibility of failure by hurrying-up too quickly – perhaps modelling the over-looking of a familiar battle-drill or accepted field practice.

Good troops should be able to accomodate a little urgency, poor ones should be penalised by their rushed, panic-made mistakes.

---

Underwear drill – all goes into the laundry basket outside the bathroom – on your way there.

The old saw is that all probabilities are 50-50– either something will happen, or it won't.

One could also argue that the Universe is entirely deterministic, but appears to have random events because we can't, don't or won't perceive the underlying universally-connected mechanisms of cause and effect that inevitably produce any apparently chaotic event. (Calvin would nod in approval, but I am not a Calvinist, so I prefer to doubt this concept.)

I any case, it is the Gambler's Fallacy to assume that any specific outcome of a random set of probabilities either will, won't, or must or mustn't occur in any given instance. Yes, a perfectly random d6 could produce a 1 over a thousand times in a row, because the very definition of randomness means that the 1 is possible for every roll. That we aren't likely to see this, or would be suspicious if we did, doesn't actually matter to the event itself. That's us making assumptions. And in a wargame, the destruction of assumptions is often hilarious.

Roll dem bones!

"Since dice are no longer relativity speaking relevant, I can just declare that I won and go home"

In the quantum multiverse every dice you roll scores every possible number simultaneously, just not in this specific instance.

However, having been trained in statistics and probabilistic modelling, most useful probabilities are based on data analysis, probability distributions and models. Not everything is a bell curve. Bookies are very, very good at calculating probabilities.

"Herr Professor Schroedinger! Your cat!"
"He rolled a …. What did he roll???"

Personally I've never liked using numeric dice in my games. I wanted to play wargames but realised what I was playing was dice games, or card games etc..
Also the dice (etc) determine, in many cases, THE outcome, which tends, because we are dealing with numbers, towards a binary, where we end up with a binary where we get a high stakes positive or not.
Yes ultimately things are binary. They happen or they don't. However in life thjngs are more likely to be on spectrum of possiblilities.

My preference is also for front loading the variables and then the use those to produce outcomes, or at least a range of outcomes.

Thanks again John.

For completeness:

UK Gov – Defence Intelligent Communicating Probability

Defence intelligence: Probability Yardstick
===========================================

1. Remote chance (0% – 5%) [5]
2. Highly unlikely (10% – 20%) [11]
3. Unlikely (25% – 35%) [11]
4. Realistic possibility (40% – 50%) [11]
5. Likely or probable (55% – 75%) [21]
6. Highly likely (80% – 90%) [11]
7. Almost certain (%95 – 100%) [6]

[76%]

Missing:

6% – 9% [4]
21% – 24% [4]
36% – 39% [4]
51% – 54% [4]
76% – 79% [4]
91% – 94% [4]

[24%]

Like my wife telling me to pick up my dirty underwear off the bedroom floor. She might as well roll a D6 to determine how many days it will take.

That may decrease the probability of staying married, however.

Stoppage: Why all the hate on Simpkins?

And considering that Kennedy pulled the air support as well as other support from the BoP forces, it's no wonder it was considered a fiasco.

Begging the question that it was the smart thing to do in the first place. Which is a proper discussion for another different thread.

Like my wife telling me to pick up my dirty underwear off the bedroom floor. She might as well roll a D6 to determine how many days it will take.

That may decrease the probability of staying married, however.

Fortunately for me, she has a -5 die roll modifier for the last 32 years.

Woldhag

Perhaps "Scientific American: Why Probability Probably Doesn't Exist" article is highly improbable.

Gamesman6, you need to read Clausewitz.

"There is no human practice that is so constantly and so pervasively in touch with chance as war." Clausewitz, Vom Krieg.

My son when he was in high school, conducted dice rolls with 20 different colored dice for a science fair. He won local regionals and placed at the State level. Won a laptop as the Armed Forces award. He explained all of the variables such as, rounded edges, glitter, air bubbles, different material(swirls.
I was proud of him.

Rmaker
I have.

You mistake what I mean as a rejection of chance, it's a rejection of conflating the chance provided by dice and in this topic the probabolities we create to allow dice to function, and the chance that exists in life and war specifically.

Any interatiin between contesting wills, has chance.
Again my issue is with numeric dice and the probabilities that get attached to the mechanisms.

PS I hate the concept of "activation" – smacks of someone back at brigade with a bubble-gum chart of the battle.

Me too.

We typically did a 20-mile forced march in 5 hours at 4mph. The platoon would start stringing out, but every hour we had a rest stop where they caught up.

However, we could have done it at 4.5mph which would have resulted in more stringing out and some not making the 20 miles. You get the idea.

Sonewall Jacksons Foot Cavalry could march 30 miles a day but how many made it that far as the average was 15 miles a day(?).

Have you ever played a game where you charged and came up 1/2 inch of the enemy? Bitch, right. Why not let the unit get that extra movement but at the cost of friction being less efficient or combat in the charge?

The US WWII War Games manual uses deterministic not die rolls for causalities.

Wolfhag

I like that idea that charges don't randomly fail, but rather lose some degree of effectiveness if they "overdo" the charge range/

Before the event, things are probabilistic, there is a chance that it can go either way. After the event it has become binary, it has either happened or has not happened. Dice are a mechanism for converting one to the other.

"Failing" charges bugged me especially when the player can premeasure.
Though it's exacerbated in IGYG rules.

Probability if something, and if something we've accurately quantified, moving from that to an outcome

Again I don't think numeric dice achieve the goal well as it shifts the focus of the interaction.

Let's complicate Wargame rules. Let's drive people who like simple rules (like me!) away. They have no business even thinking about engaging in simulations.

John, you need not worry, you can separate wargames rules from military simulations. Wargames rules as as complicated as you want to make them and are 100% accurate because you say they are. If you want to delude yourself [no names no pack drill] that these rules are a military simulation go ahead [I'm pretty sure you don't], but be aware that like all military simulations they are probably massively inaccurate in some key areas.

The issue goes back to a philosophical point first made by David Hume in the early 1700's. Scientific logic is based on data and is only as accurate as the data; however calculating the accuracy of the data is always limited by the unknown unknown, the inaccuracy you are unaware of, therefore at best it is a guess. This applies to everything you think you know, not just those things you think you can quantitate. It is one of the most important philosophical points ever made in the history of man, but it is not a popular one [since it undermines every person who ever said "I know", or "the answer is"] so most people ignore it.

Wargame rules tend to be too complicated already, in large part because of abstracted ways to "solve" the perceived need and reliance of probabilities.
Getting rid of or altering that relationship can simplify process and the rules.

Regarding David Hume's observation:

When I was learning about computers back in high school, very early on I learned the term GIGO, or Garbage In, Garbage Out— which in computing circles meant that your results (output) of a computer program will always only be as good as the input (data) given the program. But that also meant that the program itself could be "garbage"— i.e. that even a perfectly debugged program that appeared to perform as expected for its purpose might actually be flawed in its computations, so that certain known and accurate data might produce known and expected results, but other data (even if accurate) might produce faulty results even if they appeared to be what was expected. (And, of course, the expectations could be wrong themselves— I think this plagues a lot of "computer model" science today, with the machines programmed to produce the result the scientist expects (i.e. "confirms" his theory), without any real examination as to whether the expectation is at all the correct one.

So the same is true of games as well— the mechanics may well produce things that "look right," but if you're expecting the thing to be an accurate simulation of war, you're probably way off the mark.

My take is that most rules use flawed value sets to base mechanics on. Let's face it, we want to know the results of an issue and move on from there. Does it really matter if the 41st regiments was performing the mechanics of firing in accordance with stated regulations but failed to hit the broad side of a barn? Of course the result is what we are after. You can make excuses/reason why the results was not as expected but as the leaders know, them's da cards that were dealt and ya gotta play the best you can with them. TRULY, crafting a successful plan of action is up to the quality of the unit's leaders and their ability get performance out of the men. How many games stop to account for the efforts of the actions taken by the unit's leadership to mantain/regain control and better coordinate efforts of the men within the confines of the turn? DO we really need to know what methods were used/ not used? NO. We only need to determine how successful they were in keeping the unit functioning. In that determination, what the heck do the numbers (how many men left as one example) are there have to do with unit functioning? It's the effects of their actions upon the enemy unit that is important; is the unit holding their ground?; do we have enough to attempt another try at a charge?; Who really gives two cents that Smith's rifle exploded with 6 loads in the barrel because the Corporal didn't pay close enough attention and stop him from doing that? Knowing that that tidbit of trivia will not change the fact that one less musket is in the firing line and Smith's comrades on either side of Smith may need the corporal's attention to redirect the men's focus on the task at hand and mourn Smith later. THAT is a simple example of what happens in reality so why not track how efficient the unit's leadership is at their job of accomplishing the unit's assigned objective? The military is NOT a democracy! "Well I'm leaving- who's with me?" is Hollywood's idea for making of an interesting script, it's not what happens as the norm. In some periods, it was that corporal's duty to stop (or even kill) that runaway and maintain order (and therefore) control over the men. SO checking to see the results of the leaders to maintain control is in order. Unlike many systems that will say "I do that in the morale check" isn't so! In a check of the unit's leaders efforts, if successful- unit capability should be regained. If failed, nothing worse happens since we already know how bad the unit is suffering at the snapshot we test. So essentially, either the leadership succeeds (resulting in an reclaimed amount of combat efficiency) or the next turn's test may succeed. Not all tasks ALWAYS take the same amount of time to complete. Rather than to create charts to crank that suggests a best attempt to cover all bases (with a corresponding delay in game), the result is either successful or not- move on. 15 or 200 guns firing – only the results are what counts in the ability of an opposing unit to function. How many are hit is not all that material if the unit remains in place. By shifting emphasis away from numbers of casualties as to the effect upon the target unit to stay their ground (results) is the nugget we are after. The faster we can determine that the faster the game will play. Just some calling out of bad value sets forced upon us since the 60s (or maybe even before that?) Time for a change!

As usual say… how much of what we do in games is based on ideas that have belike defaults, whether they are actually useful or valid.
We use Dice as a randomiser, mainly d6 because early games had them available.
So often I read in design notes that they use d6 because they are easy to obtain..
People are shown to want something to do so dice rolling fits that needs.
But then were stuck by the limitations offered by those 6 faces.

Chuck in perceived probabilities, accurate or not, that need to be made to fit in to those limited numbers offered by the dice etc.
We're not only looking more at the out come of the system not the action we are plating, I need a 6!
But we're designing systems because of the constraints of what we think that system should be.

We end up rearranging deck chairs on the titanic… because we've got deck chairs…

Before the event, things are probabilistic, there is a chance that it can go either way.

Yes, for some actions some of the time.

At the 1:1 and skirmish levels, I believe that before the event, a leader would estimate how long it would take to accomplish an action, move from point A to B, or execute an order. You may not see it as a probability unless you are under fire, experiencing unforeseen events, or encountering obstacles. I don't think it's so much a probability but more of a timing issue. Since very few games are Time Competitive (like a real simulation) you are left mostly with probability and die roll modifiers.

If you have a squad doing fire team rushes of 5 yards, it's not a probability of how far and if they'll move other than enemy fire and suppression. In a shorter time span, there are fewer variables. When you get into minutes, hours, and days, a variety of situations come into play that will generate friction, delays, miscommunication, and mistakes.

Wargame rules tend to be too complicated already, in large part because of abstracted ways to "solve" the perceived need and reliance of probabilities.
Getting rid of or altering that relationship can simplify process and the rules.

That's what I've tried to do.

When you have a turn that is a fixed amount of time, say 30 seconds, as the designer your goal is now how do you parse the actions for each unit throughout the turn using traditional abstracted rules, mechanics, and die rolls/modifiers. It seems the majority of players have found a system that for them gives the right "feel" which is normally the goal.

Creating Friction for a turn with a set amount of time: Making your units perform above their expected performance can disorder or fatigue them. Vehicles moving at an unsafe speed can bog down or throw a track. I like letting players use historic Risk-Reward Tactical Decisions. This could be done by giving a modifier to initiative or activation but with an accuracy or movement penalty.

Stochastic Timing using historic values with some variations: If a gun's historic average reload time is 8 seconds, rolling a D10 for reload time might have a 10% chance of reloading in 7 seconds, a 60% chance in 8 seconds, a 20% chance in 9 seconds, and a 10% chance in 10 seconds. Unless the loader is interrupted or slips/trips/drops the round will eventually be reloaded unless he is killed. In 1:1 engagements, seconds count.

Deterministic: I've played simple intro games in which Russian tanks fired every 8 turns, German tanks every 6, and anti-tank guns every 4 seconds. When engaging a new target, tanks with a turret add a D6 number of turns. Assault guns with no turret add a roll of 2x D6 number of turns.

As each game turn is announced sequentially, players pause the game to shoot, observe the results, and issue their next order. Since all units and crews are active as in real life there are fewer probabilities. Stochastic Timing presents a better Fog of War as no one is sure which unit will pause the game to react, engage, or shoot.

On game turn #20 a tank engages a new target taking 11 turns, he will pause the game to shoot on turn #31 (recorded secretly) if he is still alive.

While on turn #31, he pauses the game, shoots, and observes the results, a miss. To reload and shoot again, he'll shoot on game turn #39 if he is still alive, engage a new target (roll a D6), or issue a move order. No other rules are really needed.

The problem with Deterministic is that the game will not be fair and balanced. Guns with a historically higher rate of fire will fire more often.

In one game, the German player complained that the Shermans were almost always shooting first and shooting more often, not very fair and balanced. I gave him the historical documentation and he stopped complaining. Sometimes reality sucks.

You could say timing it is IGYG where I Go before you because I'm quicker, You Go after me if you are still alive. This eliminates the initiative and unit activation rules and the probabilities associated with them.

No activations (units are always active to react and engage), no orders or spotting phase, no initiative determination, no command points, etc.

When a game turn is announced and no one pauses the game to shoot or react, the next turn is immediately called (no other rules are needed). The game moves rapidly to the next player's action with no additional rules or die rolls.

Players record their times in secret, creating an interesting Fog of War as no one knows exactly who will go next.

One problem I've seen with the probability of unit activation is that poor die rolls can result in an unhistorical outcome or a non-historical result. Say you activate on a 1-5 but four times in a row you roll a 6, it happens. The solution in that case, each attempt after the first unsuccessful roll gets a modifier for a better chance. That could somewhat represent timing.

Wolfhag

@ Dye4minis "By shifting emphasis away from numbers of casualties as to the effect upon the target unit to stay their ground (results) is the nugget we are after. The faster we can determine that the faster the game will play".

Did that, wrote the rules [for Napoleonic's] several years ago, no one will play with them [see my request for play testers and the complete lack of feedback.