Can you profit from trading if you know the next percentage change will be smaller or larger than the last one?

[Flatino]

Hi everyone!

I'm completely new to trading and have never actually traded before, so apologies in advance if this is a naive or unclear question.

I’ve been thinking about a theoretical scenario:

What if we could somehow know in advance that the percentage change in price (the intensity of the move) in the next tick or time frame will be smaller or larger than the previous one? Not the exact price itself, but just the magnitude of the change compared to the last move.

Is there any way to use that information to make a profit when trading?

I realize this isn't the same as knowing the price direction or value, but I’m really curious if this kind of insight could still offer some kind of edge.

Thanks a lot in advance for any thoughts or advice!

I'm eager to learn and appreciate any guidance.

Kind regards

Flatino

[ShaunWhite]

What if we could somehow know in advance that the percentage change in price

I realize this isn't the same as knowing the price direction or value.....

It's the same but you're also adding the expected size of the move.

To make it short, if you can predict anything at all in advance then you can make money from it. But if you're new then don't over complicate it, markets are just games of supply and demand so 'all' you have to do is figure out which way the price might move and buy or sell accordingly, on that market. If you're wrong then exit with a small loss, if you're right then try to make more than when you lose. Eg If you get it right 50%.of the time and if you make +2 on the wins and -1 on the losses then you're in profit.

Forget about when the next market might do.

[Flatino]

I don’t think we’re on the same page. Let me clarify: for example, if the price drops by 10 ticks, my prediction is that the next move will be of equal or smaller magnitude IF it goes in the same direction. That doesn’t necessarily mean the price will go up , it could drop again, but then it will be by less than 10 ticks, or it could go up by any amount. The key point is that I’m not predicting a reversal, just that a larger move in the same direction is unlikely.
But if I back it and the price drops again — even by a smaller amount — I still lose, even though my prediction was technically correct.
I hope that makes more sense now

And yes there is a way to accomplish this with astonishing hit rate but practicality of making money with it is what I am struggling with.

The price itself and the change I’m referring to come from two different statistical distributions...

Cheers and thanks

[ShaunWhite]

Let me clarify: for example, if the price drops by 10 ticks, my prediction is that the next move will be of equal or smaller magnitude in the same direction.

What timescales are you talking about, the markets are always moving so your interpretation of magnitude in a period seems to be subjective rather than objective? And are your time periods equal, do you want to know if a move in any 10s period going to be followed by a bigger/smaller move in the next 10s period. Or could that next period be 20s.

It might be easier to explain how you plan to use that knowledge because it sounds like you're looking for a way to recover a loss from a previous decision, when in fact that has gone, it's history and what you do next shouldn't be influenced by it.

[Trader724]

Let's consider the information your prediction provides. You're essentially gaining insight into the local volatility or momentum decay.

It's like knowing the wind is losing speed, useful for kite flyers perhaps, but not enough information for a sailor to set a new course.
Just knowing that the intensity of the price movement will change is not enough to generate a profitable trading signal. You need a prediction of the price direction to be able to make a trading decision with a reasonable probability of success. Your information about intensity could be valuable as an auxiliary tool for other strategies or for risk management, but not as a primary standalone trading strategy.

The main usefulness of your information about the slowing intensity of price movements is that it begins to indicate the emergence of characteristics of a market that could be favorable for trading strategies that perform better in lower volatility conditions. It's not a guarantee. Just because the intensity is decreasing doesn't automatically mean the market will become perfect for these strategies. Events can occur at any time that reintroduce volatility.

Essentially, your information acts as a preliminary signal or filter that alerts you to a possible change in the market's volatility regime, opening the door to considering strategies that thrive in such conditions.

[Flatino]

The time scale doesn’t matter it can be minutes, days, whatever. The principle stays the same no matter what. And just to be clear, there’s no subjectivity here. It’s all grounded in math and statistics.

You don’t need a degree to understand it either. In fact, it’s simple enough that even a five-year-old could follow the logic. I’ve always been fascinated by numbers—especially random ones. That’s really where this all begins. Randomness is everywhere, and it quietly shapes everything around us.

What makes this even better is that anyone, anywhere, can test it. You’ll see the same behavior every time you roll a dice while playing Monopoly with your family or just watching how prices change on the stock exchange. Also for example any true random number generator will follow this principle. It has to. That’s the nature of randomness and it’s not just chaos, it has structure.

I might share what I know and hopefully it might be beneficial for someone somewhere.

And to be honest, I really like the way Trader explained it—it’s well put and lines up perfectly with what I’ve found while trying to apply this in the way I originally intended, so far.

P.S. The distribution of goals in some sports also follows this principle if viewed from the right angle

[Fugazi]

I might share what I know


I'm all ears

[ShaunWhite]

Your inital post didn't sound like stochastic modelling but if that's what you mean then that's something which underpins most trading especially in my field which is HFT quant trading, where technical analysis in all it's various forms is commonplace.

If you're inexperienced and planning to trade manually you might struggle with your sample size to exploit it much, if you're mathsy and plan to automate then start collecting tick data asap.

[Atho55]

This is a view of a football predictor model that simply looks at the Count of Home, Draw & Away by Round then looks to indicate what to do at the next round of games. This week (no round number at the bottom) is Lay Lay Lay.

Look at the higher numbers in the data, 6`s and 7`s and you can usually see a smaller number around them. This I think is the general principle the OP is on about. If you get 7 Away wins one game round you are not likely to get 8 the next although it could happen.

Prem Lge Freq of 1X2 by Round.png

[Flatino]

We're going to play a game.

You will secretly choose two different real numbers, which we'll call A and B. These numbers can be anything you like—as long as they are real (not imaginary) and different from each other. They can be positive or negative, rational or irrational ( For example: —π, e, -300, +1000) it doesn’t matter.

Once you've picked A and B, you will randomly reveal one of them to me. I won’t know which number is still hidden, only the value of the number you've revealed. Then, you’ll ask me to guess whether the hidden number is greater than or less than the one you showed me.

Surely you will think I only have a 50:50 chance of guessing correctly. But here's where the twist comes in:

Before seeing your revealed number, I will generate my own random number, which I’ll call K. I’ll draw K from a probability distribution that covers the entire real number line—for example, a normal distribution with any mean and standard deviation I choose.

Now, once you reveal a number—say it’s A—I compare it to my random number K.

Now I will say:

If A > K, I will guess that A is the larger of the two numbers.
If A < K, I will guess that B (the hidden number) is larger.


That’s it! Just by using this method, I can tilt the odds (significantly ) above 50% in my favour.

But why does this work? There are three possible scenarios:

1. K < A & B (K is smaller than both A and B) → I have a 50% chance of being correct.
2. K > A & B (K is larger than both A and B) → I also have a 50% chance of being correct.
3. K is between A and B → This is where the magic happens: in this case, I always win, because I can correctly deduce which number is bigger.

So to sum it up:

If K is smaller than both numbers or larger than both, you’ll win about half the time in each case. But if K falls between the two numbers — which it must with some positive chance — then you always win in that scenario!!! That middle case adds extra winning probability, so your overall chance of picking the larger number ends up being greater than 50%. For example, if the two numbers are 1 and π, this method gives you a better than 57% chance of choosing the larger one.

I’ve already explained why I know this works, but now I’m really interested to see who can test it properly and what the percentages come out to. I’m confident in the logic, and I know it holds up.

I’m sure there are people out there with enough coding skills to test this — so go ahead, prove it to yourself. If you think this is just some mathematical trick or sleight of hand, write a quick simulation and see what happens. It’s not complicated, and if you really know what you’re doing, it won’t take more than a few lines of code.

Don’t just take my word for it—test it yourself. Run the numbers, simulate the scenario, and challenge the logic. I’ve laid out the reasoning, but now it’s time to see the actual results.

If you manage to disprove what I’ve explained, I’d be open to seeing your findings. The challenge is yours—let’s see who can accurately test it and what the data says.

Then we will work from here.

[Fugazi]

Just use martingale mate

[jamesedwards]

I've recreated this using excel.

K2.xlsx

I think the reason it works it because knowing whether K is greater or less than A removes a proportion of possible values from B, thus shifting the odds of a correct guess in your favour.

I see no way to transfer this to practical application in the betting world.

[Fugazi]

I've recreated this using excel.
K2.xlsx


I think the reason it works it because knowing whether K is greater or less than A removes a proportion of possible values from B, thus shifting the odds of a correct guess in your favour.

I see no way to transfer this to practical application in the betting world.

On larger spreads, if one side keeps getting bets matched then you know the true value lies closer to that side than the other could be an application. But then again, early markets best lay price is nearly always later to approach true value than best back price so maybe it won't really give any useful info

[Flatino]

Thanks for your effort, James. So it’s working. Let’s not worry about the potential application for now—I’m not saying there is one. If it were that easy, I’d probably be driving a Bentley and living in a château in the south of France.

But it’s still a really interesting observation. Now, what if I told you we could bump that prediction up to 70% in certain cases? I know, once I explain it, you’ll probably go, “Well, duh!”—because it’s actually that simple and logical.

This works when we know the exact range of possible numbers we’re choosing from. Say I ask you to pick a number between 0 and 100. We use the median—50—as the K value. If the number is smaller than 50, we guess the next will be bigger; if it’s bigger, we guess the next will be smaller. That’s it.

And here’s the beautiful part—it doesn’t matter how big the range is. Whether you’re picking from 0 to 100, 0 to 1,000, a million, or even a billion… once you know the median and use it as the K value, you’ll consistently hit around 70%. It’s clean, simple, and it just works.

You can even play around with this using dice. Roll one die—your range is 1 to 6, so the median is 3. If the number is less or equally( sorry I originally forgot to say this) than 3, guess up; if it’s more, guess down. With two dice, the range is 1 to 12, so the median is 6. Try it—it’s actually fun, and you’ll see how surprisingly well it holds up.

It’s so easy you can test it on the fly just by glancing at a few numbers—remember what I said about a 5-year-old being able to do it? Yeah, it’s really that straightforward.

Cheers

[jimibt]

flatino - enjoying this wee exercise in theoretical maths.. thanks for that.

the one thing that may be problematic with decimal odds is the fact that the ranges are not contiguous numbers, i.e.

The increments between odds vary depending on the range:

1.01 to 2.00: increments of 0.01.
2.00 to 3.00: increments of 0.02.
3.00 to 4.00: increments of 0.05.
4.00 to 6.00: increments of 0.10.
6.00 to 10.00: increments of 0.20.
10.00 to 20.00: increments of 0.50.
20.00 to 30.00: increments of 1.00.
30.00 to 50.00: increments of 2.00.
50.00 to 100.00: increments of 5.00.
100.00 to 1000.00: increments of 10.00

this may or may not have an impact, but you can see that the leap from 1.01->2.00 has a much larger delta range than the higher numbers where for example the increments on 100->1000 is 10. that makes a median number quite difficult as the ranges above and below the median will be apples and oranges.

just an observation