Why are you so fixed on creating a quadratic equation? You don't have enough points to say, "I can fit a porabola that accurately goes through these points."
If you take the time to accurately measure xp, you can get data points for every level. So if you were to calculate the xp required to level for every five levels, you would have a pretty accurate set of data points. If you put those points onto a graph, you will see a general shape. That shape will tell you which type of equation to start with.
Keep in mind that OTM could be using a really complex equation that could be replicated accurately enough with a simpler equation.
Well to be technical they don't even need to use an equation. Rather a lookup table could be used which is the more likely solution to what is coded - it was generated perhaps from a single function or even piece wise functions. It could change at any time also and tho it seems unlikely could be class dependent. Or it was simply read off fortune cookie lotto numbers. we may never know.
Never underestimate the power of three data points to predict the next 207. Or was that overestimate? The only thing I'm sure of is it never pays to drink and derive.
Firstly, there are many equations that can fit 3 data points that close.
Secondly, three points can tell a lot, but in this case, not too much.
Thirdly, a table represents the equation's variables and and their relations.
Fourthly, the game is set up (and I think it was an admin who said this) so that the time to level increases linearly. We know that xp per mob rises slowely as you level and mobs don't take that much longer to kill as you level. This means that the rate at which xp required to level is related to level. A rate of change is also known as a derivative. If there is a derivative, there is an original equation.
So yes, there is a table. That table includes those three points as well as other points. However, you cannot have a table of this sort without an equation.
There are an infinite number of equations that can pass through 3 points. 3 points, without other knowledge, say nothing about the actual function, if in fact it can be defined as one. I could talk about the nyquist sampling period, but it's a waste of time. The actual table is probably stored as an array variable, with no need whatsoever to conform to a linear formula, which was my point. Not that you can't fit one and make it work if you have enough data lols. You can easily have a table that does not fit any equation I have coded similar things for approximating thermistors and other sensors - custom calibration to unwarp the data more accurately and with far fewer CPU cycles than any sane function can provide.
It's entirely possible that each successive level was just computed by multiplying by 40% and rounding to the nearest 100xp. Rinse and repeat and no continuous function will line up with this simple procedure.
Also to be technical again, just because there is a derivative does not mean there is a function that can be solved for in open form. Integration is not always easy. Plenty exist where your only option is to pick a starting point and numerically extrapolate the answer. In the linear case it is quite trivial obviously.