the Mark98 claim

Jay : Markybot : Mark98

The Mark98 FX claim pays off depending on Markybot’s score after a year of operation. Well, technically it pays off depending on player 14’s score, but player 14, registered to me, is currently traded by the robot Markybot, and I intend to keep it that way for at least the life of the claim.

This table shows how the claim’s payoff depends on the score.

score	payoff	score	payoff	score	payoff	score	payoff	score	payoff
0.750	0	0.850	20	0.950	40	1.050	60	1.150	80
0.755	1	0.855	21	0.955	41	1.055	61	1.155	81
0.760	2	0.860	22	0.960	42	1.060	62	1.160	82
0.765	3	0.865	23	0.965	43	1.065	63	1.165	83
0.770	4	0.870	24	0.970	44	1.070	64	1.170	84
0.775	5	0.875	25	0.975	45	1.075	65	1.175	85
0.780	6	0.880	26	0.980	46	1.080	66	1.180	86
0.785	7	0.885	27	0.985	47	1.085	67	1.185	87
0.790	8	0.890	28	0.990	48	1.090	68	1.190	88
0.795	9	0.895	29	0.995	49	1.095	69	1.195	89
0.800	10	0.900	30	1.000	50	1.100	70	1.200	90
0.805	11	0.905	31	1.005	51	1.105	71	1.205	91
0.810	12	0.910	32	1.010	52	1.110	72	1.210	92
0.815	13	0.915	33	1.015	53	1.115	73	1.215	93
0.820	14	0.920	34	1.020	54	1.120	74	1.220	94
0.825	15	0.925	35	1.025	55	1.125	75	1.225	95
0.830	16	0.930	36	1.030	56	1.130	76	1.230	96
0.835	17	0.935	37	1.035	57	1.135	77	1.235	97
0.840	18	0.940	38	1.040	58	1.140	78	1.240	98
0.845	19	0.945	39	1.045	59	1.145	79	1.245	99
								1.250	100

it’s hard to predict

I think the claim is interesting because it’s hard to predict Markybot’s score. If the bot loses money, I will get annoyed and work to improve its algorithms, possibly breaking a downtrend. If it earns money, other players may take note and try to copy its strategy, possibly breaking an uptrend. Nothing can be counted on for sure.

Over its first four weeks, Markybot’s score varied between about 0.970 and 1.045, a payoff difference of 15. Broad score trends are fairly slow. But note that a drop from 1.030 to 0.980 happened in one day, so occasional sudden changes are possible. I think the claim should be interesting both to fundamentals players, who think they can predict the outcome, and to traders who try to take advantage of volatility.

manipulation

It will be possible for fanatical Mark98 holders to manipulate Markybot’s score near the judgment time by pushing around the prices of claims Markybot holds positions in. Because Markybot holds an extremely diverse portfolio and avoids taking large positions in thinly-traded claims, it should be expensive to manipulate its score by much. I expect that manipulation is unlikely to be profitable. Nevertheless, the claim does allow it, and you should be aware of the possibility.

Markybot itself will not trade in this claim, at least not unless it suffers a bug, its account is broken into, or something equally dire occurs. If it does buy the claim by mistake, I intend to intervene by hand, as explained below, to sell the coupons, as quickly as I can after I’ve fixed the bug. Note that the claim definition does not require this.

human intervention

Currently, Markybot makes all its own buy and sell decisions, and I haven’t intervened. Even when it bought TWAC due to a bug in the judgment date checking, I did not step in to sell the coupons it bought by mistake. I even let it buy more, although I expect it to lose its money.

However, the Mark98 claim does not require that Markybot make all its own decisions. The claim is about the score of Markybot’s account, not the score that Markybot earns algorithmically. And of course I wrote the program, so in a sense Markybot doesn’t make its “own” decisions anyway; it makes the decisions I designed it to make. One of the ideas I want to try out in the future is giving Markybot hints about how I expect claims to be judged. I might tell it, for example, that I’m 90% sure that MLAW will be judged 0, and let it trade accordingly. I may make other interventions as well, though I don’t currently plan to.

updated 20 June 2000