Finally, a block!
We had superb luck when the rental rates are too high to bother with. One of my rentals came alive last night and has run through much of today and a second one has occasionally been alive today. Of course-- now we get bad luck so that I can lose on those again! When exactly does the variance swing my way during a rental?
Edit: Yes, I know this isn't that bad. I've only been playing with adding rented hash to my own gear for a couple weeks. The timing just annoys me... the pool is doing around 115% luck during that time and I'm showing a 20% loss on the rentals. I'm not risking a lot and I will not make any conclusions until at least a couple months. I'm renting well under market and the timing should average out to bring a profit-- It just hasn't been a good start and I'm hoping to see the better side for a change.
If you are not letting the rental run straight through, it's never building it's 5Nd and never getting you full payouts. It doesn't matter that you are getting really low rates...if it's not running then you are not earning shares and without getting full payouts, it will pretty much stay at a loss. Even if your rental came alive on a 6 or 8 block day, you will not be getting full 5Nd pay for those blocks because it was sitting there dormant beforehand so it's not going to make up a lot of ground with fractional payouts.
How do I know this? I have been running a steady rental under my second account for nearly the entire month. Some days I have to pay more than I really want to but I've noticed that once it reached it's full 5Nd, it does really well thanks to good performances overall by the kano pool. I've even been able to transfer 1 BTC over to my Host Fees wallet from my rental wallet and It's still operating in the black this month. Kano Pool has been averaging 0.00386697 per day per TH in my account this month so as long as I can keep the rental at least 3% less (less than 0.00375096) it pays all day long and full on each block.
Here is the strategy I've been using for my profitability test:
1) I always put my orders under 100% PPS rate. Right now that means 5% (2% after fees) and 8% (5% after fees). Originally I used 5-10-15 after fees, but the 10 and 15 just were not being hit enough to bother with. My thought is that even if the orders are not always on, it is their average rate that I am paying for and that average that I will paid for at the pool. Yes, that means occasional gaps will result in some shares falling off the N range and being paid less than 5 times due to luck, but an equal number should pile up within the N range and get paid 6+ times due to luck. If there are no other confounding variables (more on that in a moment), the math says it should all average out to 102% to 105% in the end if I am paying 2% to 5% below market after fees.
2) Once the orders are made I never touch them except to refill. This lets them work their way up to near the top of their price bracket during their 15 day run. An order near the top of 0.0029 is hit nearly as often as one at the bottom of 0.0030 but costs about 3% less, so it doesn't pay to adjust orders to chase the price.
Now about confounding variables-- this assume all others renting to point hash here are following a similar rational strategy. In reality that is not the case-- luck chasers exist who will rent well above market just to chase the luck when the pool is "hot". Of course, I know past luck doesn't predict future results, but it does still alter the distribution of my not renting periods. Theoretically it shouldn't matter-- they are renting for an unpredictable future-- when they win I should lose and when they lose I should win. When that happens should be totally random in the end.
That said, I have some experience on the flip side of this issue-- selling hash rather than renting it. A year ago I did a similar experiment to what I'm doing now except on the selling side to determine where to set my profit switch (p=) point for selling to WH. Like now, everything *should* have been random meaning I should have lost in profit switching as much as I gained. The reality *was* different though, and I concluded that any profit switching below +25% consistently resulted in losses for me. I never identified exactly what caused the phenomenon except to note that there was possibly a correlation with luck chasing at the pool. Maybe I just didn't let the test play out long enough to see a better result.
Another confounding variable is whether or not the rented hash is legitimate hash. Now that we have seen witholding attacks in the wild, it really raises the possibility that bad gear does hide on the rental services. If so, increases in the percentage of rented hash on the pool (correlated with times of low rental rates when my and others rentals turn on) would result in decreases in pool luck and would directly affect my rental strategy. Still, my intuition says this risk is low due to the combination of multiple probabilities involved (probability of dud hash existing and the probability of it getting assigned to a rental here). If I knew the percentage of bad gear on NH/WH it would be possible to run this through Bayes' therom and get an exact percent effect on the pool (it should be less than 1% though).
Part of my experiment is to track this carefully to make sure it isn't observable in my rental/reward data. I only have two weeks of data when I strip out the early 5-10-15 method, so obviously I can't draw any conclusions yet. So far a majority of my rental periods do fall within dips in reward count on my graph, but it is inconsistent as to whether they are on the leading or falling edge of the bad luck. Of course I'm not capturing the rental activity above my threshold, so others' 0.0030+ rentals could fill that out and I wouldn't know. I really need to be logging LRP over time, not just my own activity if I want to answer that question (but my main concern has been the viability of my strategy, not with the quality of hash available, so I haven't kept that level of detail). My rental activity was rewarded 5+ times in only 31.5% of shifts. The average reward count on my rental shifts is now 4.228. If the rentals are legitimate, I expect an average reward of 5 when the variance moves into my favor resulting in a profit. If it doesn't, then there may be an issue with the quality of rentals to consider.
Anyone see any flaws?
