A slow but token-efficient sort:

-- 35-token bubblesort, by pancelor
function sort(arr)
  for i=1,#arr do
    for j=i,#arr do
      if arr[j] < arr[i] then
        add(arr,deli(arr,j),i) --slow swap
      end
    end
  end
end

I did a brief speed analysis and this function seems reasonable to use for arrays up to around 50 or 100 elements, depending on how much CPU you have available to burn.

speed analysis:

cls()
function _draw()
arr={
--20,5,8,3,7,4,1,9,2,-30,
--20,5,8,3,7,4,1,9,2,-30,
--20,5,8,3,7,4,1,9,2,-30,
--20,5,8,3,7,4,1,9,2,-30,
20,5,8,3,7,4,1,9,2,-30,
}
sort(arr)
end

I found a sorting implementation that's 1 token cheaper (34 tokens), and slightly better cpu wise - based on this simple and surprising sorting algorithm: https://arxiv.org/pdf/2110.01111.pdf

function sort(arr)
 for i,arri in inext,arr do
  for j,arrj in inext, arr do
   if arri<arrj then
    arri,arr[i],arr[j]=arrj,arrj,arri
   end
  end
 end
end

complexity is clearly n^2, and also a bit better cpu wise on your tests (for 200 elements, 0.6 compared to 0.8 with yours)

pretty stoked to find a use for this niche algorithm, but i wouldn't be surprised if it's possible to do better token wise, perhaps with some add/del tricks like your alg does

oh, I recognized that sort before I clicked the link! I love it, it's so funny.

the add/deli trick doesn't work as-is unfortunately -- it requires i<=j. that's why I didn't look into using your sort, but clearly I should have -- nice token savings!