Hello, I was going to write this as a reply for some other topic… But I guess it should be here, more easily found. This is what I believe…
Apparently, there seems to be a relationship between words learned and points, ~1000 for every 1 word learned, I assume. So if a person has 20,000 words learned he’s points should be around ~20,000,000. If the discrepancy is a little high you should consider his Learn - Review ratio and extrapolate around that, if you guys (Memrise Team) keep that kind of information around of course.
Anyways, if the discrepancy is too high, e.g. you have only ~8.000 words learned, there is no way in hell you can be 308.000.000, only by underhanded means.
Hope this helps in some way, I love the community. Sorry if I seem to underestimate your (Memrise Team) understanding, that’s not my intention, I understand you guys have a lot to do with few resources, given the great business model and identity of Memrise.
If this would somehow make some cheaters go away or get their true points worth it would make my day. Anyhow, I use this algorithm to evaluate the true position of my fellow contenders on leaderboards and estimate if they are cheating or not with a quite high success rate, I believe.
Given that I already suspected that some users were cheating and then found corroborating evidence on forums, like unusual point gains reported by other users.
What do you guys think? Does my thought process make sense?
Update 1: Thanks for the replies guys, apparently there are some circumstances in which my train of thought might be flawed (when a person repeats a course, etc). Anyways, I guess it is actually not something to stress over anyways. The ones cheating are only cheating themselves. We should just do our thing and worry about the results that matter, i.e. learning.
Update 2: Aparently, this doesn’t apply always, but when I tried restarting and quitting a given course my word count did not go down in neither cases - and it did go up after relearning the words, implying consistency, ~1000 points per word (more if the user reviews more, depending can even double or triple, more than that is quite unlikely) points per word - but I guess that wans’t always the case. Anyhow, it’s fun to try to understand things.
Update 3: I’ve come up with a userscript to hide people which I find suspicious on the Global Leaderboard. Feel free to modify it to your needs, like if you weren’t going to do it anyways. 
Update 4: Now the userscript works on every Leaderboard, last I checked.
// ==UserScript==
// @name Memrise Leaderboard Cleaner Plus 3000
// @namespace http://localhost
// @version 1.0
// @description Hide specific users on the leaderboards.
// @match http://www.memrise.com/*
// ==/UserScript==
// Users you wish to hide.
var blacklist = [
'paperopunico',
'extremedream',
'HFGW'
];
$(function(){
m.render = (function() {
var cached_function = m.render;
var reorder = function(element) {
var selector;
if (element.className == 'leaderboard-container wide')
selector = 'span.row-pic > strong';
else if (element.className == 'span3 leaderboard-container')
selector = 'span.row-username > strong';
$(arguments[0]).find(selector).each(function(index) {
$(this).html(index+1);
});
};
var remove = function(element, blacklist) {
$.each(blacklist, function(key, value) {
$(element).find('span.row-username > span:contains("' + value + '")').closest('li.leaderboard-row').remove();
});
reorder(element);
};
return function() {
result = cached_function.apply(this, arguments);
if (arguments[0].className == 'leaderboard-container wide' || arguments[0].className == 'span3 leaderboard-container')
remove(arguments[0], blacklist);
return result;
};
}());
});