MathGirlSam Posted March 3 Posted March 3 (edited) This WIP guide will serve as a place for me to post analysis on the PokeMMO economy. Over time I hope to collect a large set of economic data that can be used in forecasting. I'm learning about time series analysis and econometrics as I go so there are bound to be the occasional mistakes. I THIS PROJECT IS ON PAUSE WHILE I UPDATE THE FORECASTING MODEL Pokemon Market Indices Are Under Development Because I'm Not Happy With Their Current State Long Term Trends Spoiler Volatility Graphs Spoiler Short Term Data Will Be Posted (Hopefully) Every Day What other data can we analyze? Spoiler In the future if I can get my hands on daily player count data I can begin to predict things like the money supply and start doing analysis on inflation and deflation in the economy. I am also wanting to try out different forecasting methods. If anyone has any requests feel free to let me know. Thank You So Much To The People At PokeMMO HUB 🙂 Edited 20 hours ago by MathGirlSam Overhaul 2 Update 0 CALLMESENSEIX, NoGoodLife, LyanDawn and 12 others 11 1 2 1
Coppernikus Posted March 3 Posted March 3 Good work Friend! im ngl there is something genuinely funny about bringing these economic adult life issues into a Pokemon game 😂
MathGirlSam Posted March 6 Author Posted March 6 (edited) March 6th Economic Update Spoiler The supply of Everstones has jumped causing a downturn in the short-term market. This will likely make breeding easier, causing a downturn in the price of good natured Pokemon. I was previously wrong about the Leppa Berry Market. The ARMA model seems to be holding and we are seeing a stabilization in price over the past few days. This is in spite of the jump in supply. My guess is that inflation is counteracting the supply increase and holding the price steady. Whether or not this will hold in the longer term has yet to be seen. At this point I'm able to extract Item data with ease (Thanks again to PokeMMO Hub) so if anyone has requests for any other forecasts just let me know. I have a decent to-do pile in regards to where I want to take this project so stay tuned. 🙂 Edited March 7 by MathGirlSam CALLMESENSEIX 1
ColesNightfill Posted March 7 Posted March 7 I love this project and im glad the data we can provide you is going towards something creative and productive that can spark discussion in the pokemmo economy. Really love the economic modelling and analysis, though, we must keep in mind that the market is very prone to manipulation, malicious or not, such as official events, unofficial clan events (like shiny wars), or organised buyouts. This can make things seem unpredictable or not follow a long term trend/attitude, as the situation can change quickly, which, i would say, would make it hard to draw any economic conclusions. Though I guess this would be more common in low quantity valuable items, such as vanities. The jump in leppa berry supply is very surprising and significant, without seeing much in price moving, which is interesting (this is something you discussed). Looking forward to more!
MathGirlSam Posted March 7 Author Posted March 7 (edited) March 7th Economic Update Spoiler The DES Forecast predicts stable prices for Everstones, Leppa Berries, and Heartscales. We will see if this holds up over the weekend. Today I'd like to mostly highlight some issues with my methodology for the sake of clarity. One problem is that I'm using data that isn't necessarily evenly spaced in time, which requires some extra fiddling to predict. I haven't corrected for this yet so take forecasts with a grain of salt. My current plan is to write a small program to detect missing data points then interpolate linearly to fill in the gaps. Another issue has to do with if the data is seasonal and/or stationary. This can have an impact on forecasts, and while I have been keeping a soft eye on it, I haven't paid it the attention it deserves. At the moment I just glance at a Fast Fourier Transform and Autocorrellation plot of the data. In the future I would like to switch to doing an ARMIA forecast, but as of now I don't feel comfortable enough in my knowledge of that method. I'm going to keep reading and I'll continue to post updates while I learn. 🙂 Edited March 8 by MathGirlSam Jotarorandom 1
MathGirlSam Posted March 8 Author Posted March 8 March 8th Economic Update Spoiler The influx of players over the weekend has caused some disturbances in the supply of the three item markets I am focusing on. Leppa Berries have experienced a drop in supply, while Everstones and Heartscales have seen a jump up in their supplies. If I had to guess this is because of weekend players farming Everstones and Heartscales, then spending that money on Leppa Berries to train their Pokemon and/or shiny hunt. The prices of Everstones and Heartscales is seeing a slight trend downward at the current moment, while Leppa Berries are seeing an increase in price.
MathGirlSam Posted March 9 Author Posted March 9 March 9th Economic Update Spoiler Leppa Berry supply continues to drop off as the price slowly increases. Everstones and Heartscales show the opposite, with prices slowly decreasing. Tomorrow or the day after I'd like to post about the price elasticity of Leppa Berries, as I think that will be important data moving forward. For those who don't know this is just a measure of how much a change in the supply of a commodity changes the price. I'm still working on developing market indicies for the pokemon market, but the data are lacking which makes it difficult. Currently I'm wanting to measure the PokeYen per IV ratio, but I want to come up with a solid algorithm for estimating this. This ratio changes significantly between the highest end Pokemon and the lowest end. I'll keep working on it and keep you updated. 🙂 CALLMESENSEIX, Panotxa and Jotarorandom 2 1
MathGirlSam Posted March 10 Author Posted March 10 March 10th Economic Update Spoiler Heartscales and Everstones are seeing their supply level out, as their prices continue to trend downwards. The supply of Leppa Berries is still dropping, and likewise their price is rising. First I'd like to say that I noticed an oversight in my forecasting method that I haven't gotten a chance to fix yet. Essentially the forecast isn't the most optimal it can be. I've been reading a lot over the past few days on the topic of time series forecasting so a better forecast should be coming soon. This projects scope is continuing to grow as I'm now realizing I may have to program a algorithm for solving a non-linear least squares problem myself, instead of just relying on some python library. There are plenty of algorithms out there, and I have quite a bit of experience implementing things like this so I'm not worried about the amount of work this will take. That being said it will take time. While I'm working on it I'll continue to post forecasts using the current method, but take things with a grain of salt. 🙂 gbwead, DoubleJ and Jotarorandom 2 1
gbwead Posted March 11 Posted March 11 On 3/2/2025 at 10:27 PM, MathGirlSam said: If anyone has any requests feel free to let me know. Would Battle Point Exclusive items be interesting to look at? I'm talking about Choice Band, Spec, Scarf, Assault Vest, etc.
MathGirlSam Posted March 11 Author Posted March 11 March 11th Economic Update Standard Graphs Spoiler BP ITEM GRAPHS Spoiler What Changed About the Update? Spoiler I worked on fixing the issue with missing data points. The method I used accounts for differences in time-steps so hopefully this is a bit more accurate. You'll notice that I've added grids to the graph. I find it helps me see whats going on better. Also the red dotted line is labeled, because it means something now. That is the forecast at a given time in the dotted red, and then the actual price in blue. The dark red line is the forecast for the next 12 hours. I've also included the Pearson Correlation Coefficient between price and supply for each graph. I spent the night implementing a Nelder-Mead algorithm in python and applying it to optimize a Holt-Wright Method. I'm working based off the a paper that I've listed below. Right now I'm just using section 2 of the paper, and I haven't bothered to used the improved version yet. I chose the Nelder-Mead algorithm because I had never heard of it before and it didn't involve computing Jacobians so I didn't have to break out the fortran. If anyone has any requests or questions feel free to ask 🙂 Hanzák, T. (2008). Improved Holt Method for Irregular Time Series. https://www.semanticscholar.org/paper/Improved-Holt-Metho-d-for-Irregular-Time-Series-Hanzák/18a7efd889966397bc819172f832c692172450f6 gbwead, Coppernikus, Jotarorandom and 2 others 3 2
MathGirlSam Posted March 12 Author Posted March 12 March 12th Economic Update Summary Spoiler The market appears to be entering a period of relative stability, as volatility drops across the board. My guess is that a decrease in player count due to the end of event season is resulting in a drop off in the total number of market orders, causing less deviation from the moving average prices. Standard Graphs Spoiler Volatility Graphs Spoiler What Changed About the Update? Spoiler I managed to put together a measure of volatility, though the spelling of that word seems to elude me. I took exponential moving averages that look back different amounts of time then took the root mean squared deviation from these averages. From this we get longer term and shorter term volatilizes. The higher the value of P, the longer term the volatility. I'm next going to focus on implementing seasonality into the forecasting model. This should allow for longer term and more interesting forecasts. If anyone has any requests or questions feel free to ask 🙂 gbwead and Jotarorandom 1 1
MathGirlSam Posted March 13 Author Posted March 13 March 13th Economic Update Summary Spoiler Nothing significant to report. The market continues to be stable. For clarity what I mean by stability is deviation from a moving average. This is at a 6 month low at the current moment. It will likely rise again sometime, most likely due to an in game event or a particularly busy weekend. I spent a few hours working on seasonality and a Fortran implementation of a Hidden Markov Model, but I'm not happy with the results. I now want to begin to implement a neural network with backpropagation in C++ as a supplemental forecast to the Holt-Wright method, which hopefully will provide better short term results. I'm looking to replicate what I've read in the paper cited below. Beyond this I want to look into non-linear correlation measures, as I read an interesting stack-exchange post about the subject, and I want to do Principle Component Analysis on a covariance matrix, also based on a stack-exchange post. Putting this together I may be able to start piecing together a larger multivariable model of the economy. This is probably a decent ways in the future, but I'm going to keep at it. Paper about NN for Forecasting Soenandi, Iwan & Hayat, Cynthia. (2019). A Comparison of Forecasting Building Material Inventory between Backpropagation Neural Network And Arima. IOP Conference Series: Materials Science and Engineering. 528. 012044. 10.1088/1757-899X/528/1/012044. Post About Correlation (Very Interesting Read) https://stats.stackexchange.com/questions/179511/why-zero-correlation-does-not-necessarily-imply-independence Post About PCA https://stats.stackexchange.com/questions/158281/can-pca-be-applied-for-time-series-data Standard Graphs Spoiler If anyone has any requests or questions feel free to ask 🙂 Jotarorandom 1
MathGirlSam Posted March 14 Author Posted March 14 March 14th Economic Update Summary Spoiler Leppa Berries are seeing the start of some weekend volatility, with a spike over the last 24 hours. Everstones and Assault Gear likewise spiked. Heartscales are holding steady below average, so they will be an item to keep an eye on. Moving forward I've included an exponential moving average on the graphs along with the change in volume of supply. Using the forecast of the average along with the forecast of the actual price should lead to more accurate forecasts. Something to note is that the price should hover around the average. I've started working on some more complicated models, like the neural network, but right now I'm thinking of turning my focus to correlation and covariance between different prices, and building a larger model of the economy. Standard Graphs Spoiler If anyone has any requests or questions feel free to ask 🙂 Panotxa, Jotarorandom and gbwead 2 1
MathGirlSam Posted Sunday at 02:12 AM Author Posted Sunday at 02:12 AM (edited) March 15th Economic Update Summary Spoiler Volatility continues to drop while prices trend downwards. In contrast to what I've said before, I now think that this is due to a high player count. With more players the market should theoretically become more efficient, leading to a stabilization of prices, which seems to be what we are seeing now. Unless there is an upset to the status quo I don't think we will see the market change much. Choice specs and choice scarfs have seen a huge uptick in supply recently. While specs hold steady it seems the price of scarfs looks to be falling. I've added confidence intervals to the graph. These assume that the price follows a normal distribution about the moving average. I'm working on correlation and covariance stuff right now, so that should be out in the next few days. Standard Graphs Spoiler If anyone has any requests or questions feel free to ask 🙂 Edited Sunday at 02:14 AM by MathGirlSam Panotxa 1
MathGirlSam Posted Monday at 01:50 AM Author Posted Monday at 01:50 AM March 16th Economic Update Summary Spoiler Not much to report. Everything seems to be holding steady. Correlation and Covariance stuff should be posted in the next couple days. Graphs Spoiler gbwead and Jotarorandom 1 1
bobliu Posted Monday at 11:33 AM Posted Monday at 11:33 AM Good work! This post is really interesting gbwead 1
MathGirlSam Posted Tuesday at 12:44 AM Author Posted Tuesday at 12:44 AM March 17th Economic Update Summary Spoiler Over the last day I fell down a rabbit hole reading about statistics. Correlation has been put on hold for now while I try to nail down some details. My focus now is on providing a better forecast using a stochastic volatility model. This is definitely going to take a bit to get done because I am deep in the weeds learning about stochastic calculus and quantitative finance. Because the market is so stable I decided to zoom out a bit with the graphs. Spoiler Jotarorandom 1
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now